Optical Pulse Wave Signal Analysis for Determination of Early Arterial Ageing in Diabetic Patients

Transcription

1 THESIS ON NATURAL AND EXACT SCIENCES B165 Optical Pulse Wave Signal Analysis for Determination of Early Arterial Ageing in Diabetic Patients KRISTJAN PILT

2 TALLINN UNIVERSITY OF TECHNOLOGY Technomedicum Department of Biomedical Engineering Dissertation was accepted for the defense of the degree of Doctor of Philosophy (in Biomedical Technology) on December 3, Supervisor: Professor Kalju Meigas, PhD, Department of Biomedical Engineering, Technomedicum, Tallinn University of Technology, Estonia Co-supervisor: Professor Margus Viigimaa, MD, PhD, Department of Biomedical Engineering, Technomedicum, Tallinn University of Technology, Estonia Reviewed by: Professor Emeritus Hiie Hinrikus, DSc, Department of Biomedical Engineering, Technomedicum, Tallinn University of Technology, Estonia Opponents: Professor Göran Salerud, PhD, Department of Biomedical Engineering, Linköping University, Sweden Associate Professor Miklós Illyés, MD, PhD, Heart Institute, Faculty of Medicine, University of Pécs, Hungary Defence of the thesis: January 24, 2014, Tallinn, Estonia Declaration: Hereby I declare that this doctoral thesis, my original investigation and achievement, submitted for the doctoral degree at Tallinn University of Technology has not been submitted for any academic degree. /Kristjan Pilt/ Copyright: Kristjan Pilt, 2014 ISSN ISBN (publication) ISBN (PDF)

3 LOODUS- JA TÄPPISTEADUSED B165 Optilise pulsilaine signaali analüüs arterite varase vananemise määramiseks diabeedihaigetel KRISTJAN PILT

4

5 CONTENTS List of Publications... 6 Author s Contribution to the Publications... 6 Approbation... 7 Introduction... 8 Acknowledgements Abbreviations Anatomy of arteries and pathogenesis of artherosclerosis Anatomy of arteries Pathogenesis of atherosclerosis Premature arterial stiffness in diabetes patients and importance of its early detection Non-invasive methods for evaluation of arterial stiffness Direct arterial stiffness estimation Pulse wave velocity and pulse transit time Pulse waveform analysis Photoplethysmographic signal and waveform analysis Experimental studies Experimental measurement complex (Publication I) Noise suppression in photoplethysmographic signal for PTT estimation (Publication II) Second derivative photoplethysmographic signal analysis algorithm (Publication III) Forehead photoplethysmographic waveform indices for cardiovascular ageing estimation (Publications IV) Finger photoplethysmographic waveform index for discrimination of subjects with higher arterial stiffness (Publications V) Conclusions References Author s publications Kokkuvõte Abstract PUBLICATIONS Publication I Publication II Publication III Publication IV Publication V ELULOOKIRJELDUS CURRICULUM VITAE

6 List of Publications I Pilt K, Meigas K, Viigimaa M, Temitski K, Kaik J (2010) An experimental measurement complex for probable estimation of arterial stiffness, In: Proceedings of 30th Annual International Conference of the IEEE EMBS, Buenos Aires, Argentina, August 31 September 4, (DOI: /IEMBS ). II Pilt K, Meigas K, Ferenets R, Kaik J (2010) Photoplethysmographic signal processing using adaptive sum comb filter for pulse delay measurement, Estonian Journal of Engineering, 16: (DOI: /eng ). III Pilt K, Ferenets R, Meigas K, Lindberg L-G, Temitski K, Viigimaa M (2013) New photoplethysmographic signal analysis algorithm for arterial stiffness estimation, The Scientific World Journal, vol. 2013, Article ID , 9 pages (DOI: /2013/169035). IV Pilt K, Meigas K, Temitski K, Viigimaa M (2012) Second derivative analysis of forehead photoplethysmographic signal in healthy volunteers and diabetes patients, In: IFMBE Proceedings of World Congress on Medical Physics and Biomedical Engineering, Beijing, China, May 26-31, (DOI: / _109). V Pilt K, Meigas K, Ferenets R, Temitski K and Viigimaa M (2013) Photoplethysmographic signal waveform index for detection of increased arterial stiffness, Manuscript submitted Author s Contribution to the Publications In publication I the author built the PPG and piezoelectric signal modules and compiled the devices together into one working unit. In addition, the author carried out the experiments and data analysis and developed the concept of the LabView recording program. In publication II the author developed the algorithm of adaptive comb filter and implemented it in MATLAB. In addition, the author carried out all the experiments, simulations, and data analysis. In publication III the author developed the SDPPG signal analysis algorithm, processed all the signals and carried out data analysis. In addition, the author participated in planning of the experiments and conducted 75% of the signal recordings from the subjects in cooperation with other members of the group. In publication IV the author processed all the signals, carried out data analysis, and conducted 90% of the experiments in cooperation with other members of the group. In publication V the author processed all the signals and carried out data analysis. 6

7 Approbation Method and device for long term variability monitoring of cardiovascular parameters based on a registered electrocardiograph and pulse wave signals; Priority number: P ; Priority date: th Mediterranean Conference of the Medical and Biological Engineering and Computing, Ljubljana, Slovenia, June 26-30, th Annual International Conference of the IEEE EMBS, Lyon, France, August 23-26, th Nordic-Baltic Conference on Biomedical Engineering and Medical Physics, Riga, Latvia, June 16-20, th Annual International Conference of the IEEE EMBS, Vancouver, Canada, August 20-24, th Biennial Baltic Electronics Conference, Tallinn, Estonia, October 6-8, st Annual International Conference of the IEEE EMBS, Minneapolis, USA, September 2-6, th International Congress of the Medical Physics and Biomedical Engineering, Munich, Germany, September 7-12, th Mediterranean Conference on Medical and Biological Engineering and Computing, Chalkidiki, Greece, May 27-30, nd Annual International Conference of the IEEE EMBS, Buenos Aires, Argentina, August 31 September 4, th Biennial Baltic Electronics Conference, Tallinn, Estonia, October 4-6, th Nordic-Baltic Conference on Biomedical Engineering and Medical Physics, Aalborg, Denmark, June 14-17, th European Conference of the IFMBE, Budapest, Hungary, September 14-18, World Congress on Medical Physics and Biomedical Engineering, Beijing, China, May 26-31, th Biennial Baltic Electronics Conference, Tallinn, Estonia, October 3-5, International Symposium on Biomedical Engineering and Medical Physics, Riga, Latvia, October 10-12, th Annual International Conference of the IEEE EMBS, Osaka, Japan, July 3-7,

8 Introduction Cardiovascular diseases are the leading cause of death globally according to the WHO global status report (Alwan et al 2011). The number of people who die annually due to cardiovascular diseases is the largest among the causes of death. In 2008 about 17.3 million people died due to cardiovascular diseases, which constitute 30% of all global deaths. However, most of the cardiovascular diseases can be prevented through healthy lifestyle. The major behavioral risk factors of the cardiovascular disease are smoking, immoderate use of alcohol, obesity, and physical inactivity. The prevention and treatment of the cardiovascular disease is based on risk factors, which are known to cause harm to the arterial system, such as elevated blood pressure, high level of cholesterol and blood sugar, and smoking. However, a considerable number of cardiovascular events happen each year in subjects who do not qualify for treatment based on such guidelines (Herrington et al 2004). Therefore, major efforts have been placed into research to develop methods, which would enable more accurate identification of the subjects with high risk from general population. Early diagnosis and treatment of the cardiovascular disease could provide treatment benefits and reduce the cost to society (Perk et al 2012). Diabetes is not a cardiovascular disease, but more than 75% of diabetic patients die because of the causes related to atherosclerosis. The premature increase in the stiffness of the arteries in the case of diabetes patients is often explained through accelerated arterial ageing. With early diagnosis of the atherosclerosis the disease can be decelerated through treatment and lifestyle change. Therefore, it is essential to detect the early changes in the arterial walls of diabetes patients. The aortic PWV, which is an index of aortic stiffness, has already been entered to the guidelines for hypertension of the European Society of Hypertension as providing extra cardiovascular risk prediction besides classical risk factors (Mancia et al 2007). However, the atherosclerosis involves the whole cardiovascular system and changes in smaller arteries can arise even earlier. The assessment of peripheral arteries may enable earlier detection of atherosclerosis (Cohn 2006). Different devices and methods have been developed to estimate arterial stiffness (Laurent et al 2006, Woodman et al 2005). Usually the measurements are time consuming, uncomfortable for a subject and a trained operator is needed. A user independent, rapid performance and inexpensive noninvasive technique is needed to estimate arterial stiffness. The general aim of this study was to investigate the possibilities to detect the premature increase in arterial ageing using a non-invasive optical method and pulse waveform analysis. Further the aim was to compare the effectiveness of the developed method and algorithms with recognized methods used in clinical practice to differentiate between the patients with normal and higher arterial stiffness. In the first part of the study (Publication I), the measurement complex, which includes the modules for synchronous recording of optical pulse wave signals from different segments of arteries, was compiled and tested. The synchronous recording 8

9 of the signals is essential for the PTT and PWV estimation. In addition, the measurement complex enables recording of other physiological signals, which are relevant to the cardiovascular system evaluation and arterial stiffness estimation from different segments of arteries. In publication II an algorithm was developed for motion caused noise suppression in the recorded optical signal providing significant improvement of the signal quality. In the second part of study the waveform analysis of the recorded optical signals was carried out in order to determine the features of the signal characteristics for increased arterial stiffness and, consequently, to estimate the premature increase in arterial ageing. Publication III introduces the new pulse waveform analysis algorithm for the finger signal to differentiate between the subjects with normal and increased arterial stiffness. In publication IV the optical signal from the forehead was investigated by using the algorithm and optical waveform index developed in publication III to differentiate subjects with higher arterial stiffness. The aim of publication V was to compare the effectiveness of the developed optical waveform index and the arterial stiffness index used in commercial clinical devices to differentiate patients with increased arterial stiffness. 9

10 Acknowledgements I am sincerely thankful to everyone who has been helping and contributing to my thesis work. I am grateful to my supervisor professor Kalju Meigas and cosupervisor professor Margus Viigimaa for their guidance and support. I wish to thank Kristina Kööts who was actively participating in the preparation and conduction of experiments in hospital, all the supervised master and bachelor students for their helpful co-operation, Dr. Anu Ambos who helped in finding and convincing the patients to participate in the study, all diabetes patients and other volunteers who agreed to participate in the experiments, Rain Ferenets for the practical advice in signal processing. I would also like to thank my colleagues at the Department of Biomedical Engineering for providing a pleasant and inspiring work environment, Rain Kattai and Deniss Karai for fruitful discussions and competent technical advice, Jana Holmar for assistance during formal finalization of the thesis and great company in conferencing. I appreciate the time and inspiring positive support from Lars-Göran Lindberg during my semester in Linköping University. Most importantly, I would like to thank my parents Epp and Jaan who have believed in me, supported and encouraged in all the levels of education. Special thanks to my brother Mihkel who was always willing to participate as a volunteer in all the pilot studies and experiments. Last but not least, I would like to thank my girlfriend Jelena who has been patient and kind regardless of my ups and downs in the thesis work. The research was supported by the Estonian Science Foundation Grant no. 7506, by the Estonian targeted financing project SF s07, by the European Union through the European Regional Development Fund, European Social Fund s Doctoral Studies and Internationalization Programs DoRa and Kristjan Jaak. 10

11 Abbreviations AGI ageing index AIx augmentation index ECG electrocardiography HDL high-density lipoprotein IMT intima-media thickness LDL low-density lipoprotein LED light emitting diode MRI magnetic resonance imaging NO nitric oxide PEP pre-ejection period PPG photoplethysmography PPGAI PPG waveform augmentation index PTT pulse transit time PWV pulse wave velocity SDPPG second derivative photoplethysmography SNR signal to noise ratio WHO World Health Organization 11

12 1. Anatomy of arteries and pathogenesis of artherosclerosis 1.1. Anatomy of arteries The structure of the arterial vessel wall is heterogeneous and can be generally divided into three layers: tunica interna (intima), tunica media (media), tunica externa (externa) (Figure 1). The thickness and the structure of the layers vary depending on the type of the artery. The inner layer contributes minimally to the total thickness of the vessel and it is in contact with blood. The innermost layer of intima is endothelium, which is a thin layer of cells that covers the whole surface of the cardiovascular system. The smooth surface of the endothelial cells enables efficient blood flow by reducing surface friction. In addition, it prevents the blood coagulation inside the vessel. The next layer in intima is the basement membrane, which consists of collagen fiber. This layer provides a physical support base for epithelial cells of endothelium. The outermost layer of the intima is internal elastic lamina, which forms a thin sheet of elastic fibers around the inner layers (Tortora and Derrickson 2011). Figure 1. The anatomy of the artery [CC: By Kelvin Song, The structure of tunica media is more complex and its composition varies most among different arterial vessels. This layer is relatively thick in most vessels and it consists of smooth muscle cells, which are connected together with the net of elastin and collagen fibers (Dobrin 1999). This construction prevents damage to the wall of the arterial vessel at high transmural pressures. The main purpose of the muscular cells in the wall is to change the diameter of the vessel s lumen. Vasoconstriction (decrease in the diameter) and vasodilatation (increase in the diameter) is typically caused by parasympathetic intervention or in response to blood pressure. 12

13 The outer layer of the arterial vessel, tunica externa, is separated from tunica media by the external elastic lamina, which consists of elastin and collagen. In addition, this layer contains numerous nerves. In particular, in larger vessels this layer contains small blood vessels that supply the tissue of the vessel wall. In addition, this layer connects the vessel to the surrounding tissue (Tortora and Derrickson 2011). Arteries have normally high compliance, which means that a small increase in the transmural pressure causes the stretch in the wall without tearing. The elasticity of the arterial vessel wall is determined by several components. The passive (without using biochemical energy) elastic components of the tunica media are collagen and elastin. Collagen has higher Young s modulus (~10 8 Pa) than elastin (~ Pa) (Nichols and O'Rourke 2005). The elastic fibers of the vessel wall enable stretching more than two times from the initial length and they dominate the wall behavior at the low strain levels. Collagen starts to dominate at the higher strain levels. In addition, the layer of smooth muscle cells enables a noticeable change in the elasticity of the arterial wall. Largest arteries, such as aorta and branches of aorta, in the body are the most elastic. Those arteries have the largest diameter (>10mm) and a relatively thin wall (1/10 of the diameter) compared to the overall size of the vessel (Avolio 1980, Westerhof et al 1969). In these vessels the tunica media is dominated by elastic fibers. Those vessels help to propel the blood onward due to the high elasticity, while the aortic valve is closed and the ventricles are relaxing. The aorta stretches due to the blood ejection from heart into the cardiovascular system. Due to the stretch of the vessel, the elastic fibers store the mechanical energy (potential energy). It is followed by the recovery of the vessel diameter and the stored potential energy in the wall is converted into the kinetic energy of the blood. The muscular arteries, also called as conduit arteries (e.g. brachial, femoral, radial, etc.), contain more smooth muscle and less elastic fibers than elastic arteries. The lumen diameter of the muscular artery is between 0.1 mm up to 10 mm and the wall thickness comprises about 25% of the diameter (Avolio 1980, Westerhof et al 1969). The muscular arteries are less elastic due to the reduced amount of elastic fibers. In addition, the muscle tissue is able to maintain the state of partial contraction. This reduces the elasticity or increases the stiffness of the artery Pathogenesis of atherosclerosis Thickening and stiffening of the arteries is considered arteriosclerosis. The arteries stiffen with disease and healthy ageing. Atherosclerosis is the specific form of the arteriosclerosis, which is an endovascular inflammatory disease, resulting in a build up of a plaque (Cavalcante et al 2011). The disease is triggered by the chemical or physical damage of the endothelial cell layer. The damage can be caused by one or multiple of the following factors: physical damage or stress due to hypertension, turbulent blood flow in the branch points of arteries, free radicals (reactive oxygen species) in blood caused by smoking, high concentrations of LDL 13

14 in blood (hyperlipidemia), constantly high blood glucose levels in blood, high concentrations of homocysteine in blood (Libby 2002a). In healthy arteries the blood cells in blood do not stick to the endothelial cells. Due to the above mentioned reasons the endothelial cell layer gets damaged and mostly LDL starts to pass through the endothelial cells and accumulates in the intima. Oxidation of LDL in the arterial wall is caused by the exposure to NO, and some enzymes. Oxidized LDL is toxic and this initiates an inflammatory process. (Libby et al 2002). The cell surface adhesion molecules (VCAM-1) are produced due to the injury of the endothelial cells and oxidative stress (Li et al 1993, Huo and Ley 2001). Monocytes respond to the inflammation process and start to stick and infiltrate through endothelium cells into media (Crowther 2005). After entering the intima the monocyte interact with stimuli. Through formation of macrophages and foam cells, the lipid and end products of the cells are collected between intima and media layers of the artery. In case the process continues, the fatty streaks start to form to the artery wall and it is the first visible sign of atherosclerosis (Libby et al 2002). It can be seen in most of people by the age of 20 in the surfaces of the aorta and carotid arteries as a yellow hue (Strong et al 1992). It is part of the ageing process. In addition, smooth muscle cells migrate through the internal elastic lamina and accumulate, which leads to the built-up of an extracellular matrix in plaque and fatty streaks turn into fatty fibrous lesions. In the next stages fibrosis continues and sometimes with smooth cell death. This leads to fibrous cap formation over the lipid core. The fibrous cap consists largely of collagen and elastin. This represents the reaction by the body to heal the lesion. Formed fatty streaks may turn into atherosclerotic plaques or remain stable and even regress (Libby 2002a, 2002b). Continued plaque growth caused by an accumulation of LDL within the intima causes the external elastic membrane to expand. This compensatory luminal enlargement and wall thickening, known as an arterial remodeling, allows the vessel to maintain normal or sufficient lumen area and blood flow (Boutouyrie et al 1992, Cheng et al 2002 and Zieman et al 2005). However, as a plaque increases, the artery can no longer compensate by expanding outward and the plaque begins to foray into the lumen. Usually it occurs when plaque thickness reaches about 40% of the lumen diameter (Glagov et al 1987). Rupture of plaques may occur due to the biomechanical and hemodynamic stresses (Cloves and Berceli 2000). Plaques that contain a large lipid core and are covered by a thin fibrous cap rupture more easily (Hennerici 2004). The lipid core comes into contact with blood when a fibrous cap ruptures. From this stage the formation of a thrombus or clot starts. The thrombus may partially or totally block an artery causing a sudden reduction in blood flow. Partial blockage of the lumen may cause the symptoms of angina. Complete blockage of the vessel lasting more than two to four hours may cause an acute event, such as myocardial infarction (Naghavi et al 2003). Healing of the vessel may also take place. Plaque rupture with subsequent healing is believed to be the major mechanism by which atherosclerotic lesions 14

15 progress and narrow the lumen. Plaques that heal generally have a higher fibrotic composition than before making them more stable and less prone to future rupture (Waxman et al 2006). At early ages the atherosclerotic fatty plaques are seen mainly in the carotid and femoral arteries (Tuzcu et al 2001, Stary 2003). Atherogenic inflammation may appear in any arteries, but the disease is marginal in the upper limb, renal and other muscular arteries (Stary et al 1992). However, it has been reported that atherosclerosis and increase in the stiffening of arteries are strongly associated at various locations in the vascular tree (van Popele et al 2001). There are a number of risk factors for atherosclerosis, which include age, hypercholesterolemia, hypertension, diabetes mellitus, smoking, physical inactivity, and obesity. In addition, the biochemical risk factors of atherosclerosis are decreased HDL-cholesterol and increased LDL-cholesterol, lipoprotein, C-reactive protein, and homocystine (Smith et al 2011). 2. Premature arterial stiffness in diabetes patients and importance of its early detection Insulin is a hormone produced by the β-cells in the pancreas. The rise in blood sugar excites the secretion of insulin to the circulation. Insulin receptors in muscle and fat cells are binding insulin, which triggers the process where the glucose from the blood stream is entered into the fat and muscle cell (Guyton and Hall 2006). Type I diabetes results from a lack of insulin in blood. It is caused by the autoimmune process, which destroys the β-cells and due to that glucose cannot enter fat or muscle cells. This causes the glucose rise in blood, resulting over time with hyperglycemia. Type I diabetes is also called insulin-dependent diabetes (Guyton and Hall 2006). Type II diabetes is primarily caused by the insulin resistance, which is caused by the dysfunctional interaction between insulin and its receptor. Despite sufficient levels of insulin, glucose is not entered into fat and muscle cells. Increased glucose levels in blood lead to hyperglycemia and endothelial cell dysfunction. Type II diabetes is also called non-insulin dependent diabetes (Guyton and Hall 2006). Diabetes augments the process of atherosclerosis by a variety of mechanisms. Increased blood sugar has two effects on endothelium as part of atherosclerosis. The hyperglycemia elevates the production of free radicals, which are highly reactive molecules that cause damage and premature death of endothelial cells (Hudson et al 2005). In addition, the accumulation of advanced glycation end products in endothelial cells increases oxidative stress, which minimizes the availability of NO. Otherwise NO enables blood vessels to relax. Oxidative stress subsequently encourages the formation of fatty streaks in the artery wall (Creager et al 2003, Rask-Madsen and King 2007). Good diet and exercise brings faster blood flow through arteries contrary to diabetes. The force along the arterial walls, which is caused by fast and steady blood flow, has been reported in recent studies to protect the arteries from 15

16 atherosclerosis. It has been found that physical force can be a key player in the function which is capable of changing biochemical processes (Woo et al 2008). The progression of diabetes mellitus is closely connected with hypertension, renal disease and different forms of cardiovascular diseases. Diabetes is not a vascular disease, but more than 75% of diabetic patients die because of the causes related to atherosclerosis. However, about 70% of diabetic patients do not take it seriously that they are at high risk for the cardiovascular disease (Hurst and Lee 2003). However, with early diagnosis of the atherosclerosis the disease can be decelerated through treatment and lifestyle change. Elastin and collagen ratio determines the stiffness of the artery wall. Vascular ageing through endothelial dysfunction and elastic fiber degeneration leads to progressive arterial stiffness and atherosclerosis development. Arterial stiffness has been considered one of the risk markers of an early cardiovascular disease. It has been found that arterial stiffness is increased with diabetes mellitus type I (Llauradó et al 2012) and type II (Urbina et al 2010). Diabetic vascular disease can be explained through the accelerated ageing of the arteries. Earlier studies have reported that the calculated cardiovascular age, based on aortic stiffness changes, increases with diabetes mellitus about 8.9 years in the case of diabetes type I (Ravikumar et al 2002) and 11.1 years in diabetes type II (Cockroft and Wilkinson 2002). Higher arterial stiffness in diabetes patients through the measurement of aortic pulse wave velocity and aortic augmentation has been also confirmed in the earlier studies (Brooks et al 2001, Wilkinson et al 2000a, Cruickshank et al 2002). Hypertension is connected to the increase of arterial stiffness and is part of development of the diabetes mellitus (Sowers et al 2001). Arteries of diabetes patients compared to the healthy arteries are more exposed to the injuries caused by high blood pressure and more often rupture or occlude occurs (Williams 1999). Due to that the damage is caused often to retina, brain and kidneys (Safar et al 2004). However, the consequences can be effectively avoided through the blood pressure reduction. In extensive studies (HOPE and LIFE) it has been shown that benefit in the reduction of cardiovascular mortality was achieved by lowering the blood pressure (Yusuf et al 2000, Dahlöf et al 2002). Thus, the rapid monitoring of blood pressure in diabetes patients is strongly recommended (Solomon 2003, Rydén et al 2013). 3. Non-invasive methods for evaluation of arterial stiffness Several indices for the evaluation of the cardiovascular system have been developed. Generally, they are addressed to the assessment of arterial stiffness either systemically, regionally or locally (Pannier et al 2002). Due to several reasons the stiffness of the arterial wall is closely connected with the development and progression of cardiovascular diseases. Firstly, the pathogenesis mechanism of isolated systolic hypertension (increased systolic blood pressure) includes an increase in arterial stiffness (Cohen and Townsend 2011). Secondly, the stiffness of the arteries determines the ability of the vessel to accommodate ejected blood from the heart. Thus, increased arterial stiffness imposes a higher left ventricle afterload 16

17 (Chirinos and Segers 2010). Thirdly, the increased pulse pressure, which is connected to the increase of aortic stiffening, may damage the organs such as the kidney and the brain (O'Rourke and Safar 2005). Fourthly, different biologic processes are connected with arterial stiffness, which causes the cardiovascular diseases (Franklin 2008). Increase in arterial stiffness is caused by ageing and accelerated due to the propagation of the diseases, such as diabetes mellitus, hypertension, hypercholesterolemia, kidney disease, obesity, and smoking (Payne et al 2010). Non-invasive assessment of arterial stiffness is based on the measurement of indirect and intrinsically associated parameters, which can be related to stiffness. Generally, the methodologies can be divided into three groups: 1) direct stiffness estimation using measurements of the diameter and the distending pressure (local arterial stiffness estimation), 2) pulse wave velocity and pulse transit time (regional arterial stiffness estimation), 3) analysis of the pulse waveform Direct arterial stiffness estimation Stiffness, which is the rigidity of an object, can be defined through the resistance of the body to the applied deformation force. The stiffness of the arterial wall is expressed through the ration of the stress (mechanical stress caused by applied mechanical force on the unit area measured in N/m 2 ), which is applied from a given direction, and strain (relative deformation of the body measured in arbitrary units), which is the Young s elastic modulus (E) measured in Pascal s (Pa) (Nichols and O'Rourke 2005, Cavalcante et al 2011). Specifically, the Young s elastic modulus expresses the slope of the stress and the strain linear relationship within the Hooke s limits in case the material is linear-elastic or Hookean. Furthermore, the slope can be calculated by using any two points within the Hooke s limit range. In this way the calculated slope is called an incremental elastic modulus (E inc ) (Nichols and O'Rourke 2005). The Young s modulus is independent of the direction where the force was applied if the material is isotropic. The arterial wall is assumed to be isotropic in the calculations of most commonly used indices. However, the arterial wall is anisotropic and therefore the approximation is used for the estimation of vessel wall properties (Chirinos 2012). The arterial wall consists of a complex structure of the components that determine the stiffness of the arterial wall. The main components are elastin, collagen fibers and smooth muscle cells. The elastin and collagen fibers are passive (without using biochemical energy) elastic components whereas smooth muscle cells are actively modulating the stiffness of the arterial wall (Bank et al 1996, Zulliger et al 2004). Furthermore, the elastic moduli of elastin and collagen fibers are different. Therefore, the arterial wall is not Hookean and the relationship between the stress and the strain is with a curved shape. The Young s modulus increases with the increasing levels of stress. This means that the elastic modulus of the artery is dependent on the distending pressure and it has to be taken into account while comparing results (Nichols and O'Rourke 2005). 17

18 It should also be noted that the relationship between the distending pressure and strain is not interchangeable with the stress and strain relationship and the law of Laplace has to be taken into account (Atlas 2008). Relatively simple formulas can be derived for the calculation of the local incremental elastic modulus of the arterial wall by using the measurement of wall thickness in diastole and two measurements of the diameters at two different pressure levels. This calculation can be done under the assumptions that the artery has a perfectly circular shape, there are no deformations in the longitudinal direction and the wall of the artery is isotropic (Laurent et al 2006). In case the pressure and diameter are measured at the same location, these measurements can be related to the wall stiffness of the artery in this location. In addition, the parameters related to the local stiffness of the artery are estimated through the pressure-volume and pressure-area relations. However, these relationships are influenced not only by the stiffness of the artery, but also by the vessel geometry (Chirinos 2012). Furthermore, the relationships are not linear, but within relatively narrow pressure ranges can be considered as linear for larger arteries (Meinders and Hoeks 2004). In the case of smaller muscular arteries, the error can be larger (Reneman et al 2005). Thus, the measurements are carried out mainly on larger arteries. The compliance is the ratio between the relative changes in the volume to the relative change in the pressure and in addition to the stiffness the parameter is also dependent on the diameter and the wall thickness of the artery. The reciprocal of compliance is called elastance. The compliance values for arteries with different diameters are non-comparable. Thus, the compliance is normalized with the volume of the artery and it is called distensibility (Nichols and O'Rourke 2005). The volume of the artery is difficult to measure and in applications it is replaced by the vessel area or diameter. It has been done under assumptions that the crosssectional vessel area is circular and the volume increase of the artery is due to the radial expansion rather than due to deformations in the longitudinal direction (Chirinos 2012). The non-linear relationship between the pressure and the lumen diameter is compensated in the calculation of the stiffness index (Hirai et al 1989). The stiffness index has been used for the estimation of the mechanical properties of the artery in earlier studies (Miyaki et al 2010, O'Rourke et al 2002). The indices described above are used to estimate local arterial stiffness in a number of different studies (O'Rourke et al 2002, Pahkala et al 2013, Ciftçi et al 2013, Oguri et al 2013). The blood pressure should be measured at the location where the mechanical changes of the artery are detected. The local blood pressure can be estimated through the registration of the pressure waveform from the location of interest by using applanation tonometry (van Bortel et al 2001) or finger cuff (Boehmer 1987) and the calibration of the pressure waveform is carried out by using the measured blood pressure of the brachial or radial artery (Verbeke et al 2005). The calibration can be carried out by different techniques, including a general transfer function method. 18

19 To determine the vessel diameter in vivo at the diastole and stroke changes in diameter, any bi-dimensional ultrasound system can be used. However, the ultrasound systems, which include high-resolution linear array transducers, should be used to obtain high quality ultrasound image of the artery (Currie et al 2010, Nualnim et al 2011). With advanced image analysis software the mechanical parameters of the vessel can be determined with high precision. It should be taken into account that this kind of diameter determination needs a high degree of technical expertise and the procedure takes a long time. Thus, it is not effective to use it, for example, in large screening studies. In some studies, the diameters of deep arteries, such as the aorta, have been determined also by using the MRI system (Franquet et al 2013). However, most of the pathophysiological and pharmacological studies are carried out by the ultrasound system. Despite some of the disadvantages, today the ultrasound system is the only non-invasive method, which can be used for the determination of local arterial stiffness (Young s modulus) through the determination of the diameter and the IMT of the artery (Laurent et al 2006). The IMT determination has been used in addition to the calculation of arterial stiffness as a separate parameter to monitor the inward or outward remodeling (van der Heijden-Spek et al 2000, Bussy et al 2000, Boutouyrie et al 2000) of the arterial wall and as one risk factor of cardiovascular diseases (Molinari et al 2010) Pulse wave velocity and pulse transit time Pulse wave velocity is the speed at which the pulse wave propagates from the heart through the arterial system to the peripheral vessels. The pulse wave propagates through the arteries with the finite speed and it is not constant for the arterial system. The PWV depends on the viscoelastic properties of the artery. It has been generally accepted that the PWV is the most simple, non-invasive, reproducible, and robust method to estimate regional arterial stiffness (Laurent et al 2006). In addition, aortic PWV, which is an index of aortic stiffness, has been included to the guidelines for hypertension of the European Society of Hypertension to provide extra cardiovascular risk prediction besides classical risk factors (Mancia et al 2007). The pressure wave propagation without reflections along the uniform artery can be described through the transfer function. The transfer function is characterized by the propagation constant, which is a complex variable, as it has the magnitude and the phase. The magnitude is determined by the attenuation coefficient, which depends on the viscosity of the blood and mechanical properties of the arterial wall. The pulse wave propagation speed is finite and the propagation constant includes also the phase constant due to the viscoelasticity of the arteries. The pulse wave velocity varies with frequency, as the different harmonic components of the pulse wave are travelling with different speed along the artery, which is also known as harmonic dispersion (Li et al 1981, Li 2004, Callaghan et al 1984). 19

20 PWV dependency on the viscoelastic properties of the artery are expressed with the Moens-Korteweg equation (Korteweg 1878, Moens 1878), which has been corrected by Bergel (Bergel 1961) and modified by Hughes (Hughes et al 1979): PWV P E0 e h 2 2 r 1, (1) where E 0 is the elastic modulus of the arterial wall at the zero intravascular pressure P (measured in mmhg), is a constant that depends on the particular artery (typically between and 0.018), h is wall thickness, r is lumen radius, is blood density, and is the Poisson ratio. First, the Moens-Korteweg equation was derived under an assumption that the artery has a thin wall, which means that h/(2 r) is small. However, in order to decrease the error of that assumption, the equation was corrected as a result of a study by Bergel (Bergel 1961) and the Poisson ratio was added to the relationship, which was taken equal to 0.5. Young s modulus of the artery depends non-linearly on the applied pressure. In the earlier study Hughes et al (1979) showed that the Young s modulus of the aorta increases exponentially with an increasing intravascular pressure P. Then the incremental elastic modulus E inc is calculated as follows (Hughes et al 1979): P Einc E0 e. (2) The other relationship for the calculation of PWV was derived from the Moens- Korteweg equation by Bramwell and Hill (Bramwell and Hill 1922): PWV V dp, (3) dv where V is the volume of an artery segment, dv is the volume change due to the pressure change dp in the artery. Due to the difference in the Young s modulus and dimensions of the artery, the PWV is changed according to (1), while the pulse wave is propagating through the cardiovascular system. The Young s modulus is lower for proximal arteries and increases while moving towards periphery. In the case of healthy subjects, the PWV is around 4 up to 8 m/s in elastic arteries (asc. aorta, abd. aorta, carotid artery) and 8 up to 15 m/s in the muscular arteries, such as femoral, tibial and brachial artery (O'Rourke et al 2002, Nichols and O'Rourke 2005). In addition, the PWV depends on the blood pressure according to (1), which should be taken into consideration when comparing the results among the subjects and the different studies. The blood pressure should be measured as well under PWV measurements. 20

21 Although there are other methods for PWV estimation (Nichols and O'Rourke 2005, Zhang et al 2005, Khir et al 2001, Feng and Khir 2010), the technique widely used is based on the measurement of the distance between two arterial sites and the division with time it takes for the pulse wave to travel through the path length. The pulse propagation time is also called the pulse transit time - PTT. The pulse waves can be recorded by different techniques. The recorded waveform represents whether the pressure (Asmar et al 1995), the diameter (van der Heijden-Spek et al 2000), the flow velocity (Blacher et al 2003, Cruickshank et al 2002) or the volume changes (Nam et al 2013) in the artery. Those waveforms are all expressing the pulse wave, but the nature of each of them is different. However, it has been reported that the waveforms are in phase at the beginning of the cardiac cycle (Boutouyrie et al 2009a). Often the pressure signal has been recorded for the PWV estimation, although this technique may cause distortions in the waves if the pressure on the transducer is applied incorrectly (Boutouyrie et al 2009a, Pilt et al 2010a). The distance between the recording sites of the pulse waves is measured over the skin surface. It causes a measurement error because it is not the true length of the artery (Hamilton et al 2007). The error is smaller in the case of relatively straight arterial segments. The measurement error becomes noticeably larger when the two signal recording locations are situated far away and the artery structure is more complex, such as in the case of the carotid-femoral PWV estimation (Karamanoglu 2003). It is also difficult to compare PWV measurement results between different research laboratories as the distance can be measured by different methods (Xu 2003). The systematic overestimation of the PWV can be up to 30% in case the distance measurement is carried out transcutaneously between the signal recording sites of the carotid and the femoral artery (Rajzer et al 2008, Salvi et al 2008). To estimate the PTT, the time difference is assessed between two equiphase points on the recorded waveforms. The pulse waveform changes while propagating through the arteries. It has been explained through the influence of wave reflections appearing due to the branching of the arteries and the changes of the arterial stiffness and dimensions along the artery (Nichols and O'Rourke 2005). In addition, the arterial stiffness changes non-linearly with blood pressure, which also influences the PWV. Due to that the PWV is not constant within one cardiac cycle. The equiphase point is determined on the propagating pulse wave, which represents the velocity of the whole pulse wave. It has been found that the early region of the pulse wavefront retains its characteristics while propagating through the arterial system. Thus, it has been suggested to determine the PTT between the equiphase points on the waveforms, which are located at the foot of the raising front. Different algorithms are used for the equiphase point detection and PTT estimation between synchronously measured pulse waves (Chiu et al 1991, Kazanavicius et al 2005, Temitski et al 2012). The results may differ between 5 up to 15% depending on the algorithm used for the PTT estimation (Millasseau et al 2005). The gold standard method for the estimation of arterial stiffness is the aortic PWV, which is assessed as the carotid to femoral PWV (Mancia et al 2007). Due to the different methodologies and techniques for the distance and PTT estimation, the 21

22 carotid to femoral PWV estimation has been standardized (Boutouyrie et al 2010). Massive epidemiological studies have been conducted in order to prove the ability of the aortic PWV to predict cardiovascular events in certain patient groups and in population at large (Laurent et al 2006). The non-invasive estimation of the aortic PWV is difficult as the artery is hidden deep inside the body. However, the simultaneous recording of flow waves from the left subclavian artery and from the abdominal aorta just above its bifurcation by using the Doppler ultrasound enables the estimation of the aortic PWV (Lehmann et al 1998). This method needs qualified expert skills in ultra sound Doppler measurements. The alternative method is the estimation of the aortic PWV by using the pulse wave recordings from the carotid and femoral arteries, as those locations are closest to the aorta where the pressure wave registration can be carried out from the surface of skin. There are different devices that enable the estimation of the carotid to femoral PWV. The Complior System (Artech, Les Lilas, France) enables simultaneous recording of pressure wave signals from the carotid and femoral artery using two piezoelectric transducers. The other device is the SphygmoCor system (ArtCor, Sydney, Australia) (Wilkinson et al 1998), which has been widely used in clinical studies. The pressure waves have been recorded successively from the femoral and carotid artery by using the sensitive wide band piezoelectric sensor. Separately registered pressure waves are synchronized with the ECG signal and the PWV is estimated. The third device is PulsePen (Diatecne, Milano, Italy), which is similar to the SphygmoCor system. All the devices mentioned above require trained personnel and the reproducibility of the results depends on the skills of the operator. It should be noted that the pressure waves are travelling to the opposite direction from the heart in case the carotid to femoral PWV is estimated. Thus, for the path length calculation, the distance from the heart to the carotid artery is subtracted from the distance from the heart to the femoral artery. In addition, it has been assumed that the wave propagation speeds from the heart to the pulse wave recording sites are equal, which, however, is doubtful. The advantages and limitations of the first two devices are reviewed in Boutouyrie et al (2009b). A method that estimates the PTT from the single pulse waveform has also been proposed (Westerhof et al 2006, Qasem and Avolio 2008). The aortic pressure waveform is decomposed into forward and backward travelling pressure waves and the time difference is measured, which corresponds to 2 PTT for the given segment of the artery. By knowing the distance from the signal recording site to the reflection site, the PWV can be calculated. However, in a recent study a problem with this methodology has been reported, as the reflection site is difficult to determine and it seems to change the location through ageing (Westerhof et al 2008). A similar methodology has been commercialized in the Arteriograph device (TensioMed, Budapest, Hungary) (Baulmann et al 2008). It has been assumed that there is one major pressure wave reflection site in the bifurcation of the aorta. The forward and reflected waves are detected from the brachial artery by applying the 22

23 cuff around the upper-arm. The cuff is pressurized to the supra systolic pressure (systolic pressure + 35mmHg) and from the recorded waveform the time delay between the incident and the reflected waves is detected, which corresponds to 2 PTT. The distance is suggested to be measured between the jugulum (sternal notch) and the symphysis pubica (pubic symphysis), two characteristic anatomical points. Invasive validations of the Arteriograph for the aortic PWV and the AIx estimation have been conducted and relatively high correlations have been achieved (Horváth et al 2010). However, the working principle and validation is debated. Further investigations of the working principle of the Arteriograph are suggested in order to induce the medical society to bring in the device in clinical practice (Parati and De Buyzere 2010). In a number of recent clinical studies the Arteriograph was used for the aortic PWV and AIx estimation (Mulders et al 2012, Gaszner et al 2012, Ikonomidis et al 2013). The pulse wave propagation in other branches of the arteries has been studied in addition to the carotid-femoral PWV (Naidu et al 2005). The pulse wave starts to propagate when blood is ejected from the left ventricle to the aorta. The arrival of the pressure wave is detected from the artery of the upper or the lower limb. The pressure waves in the arteries were detected by using the pressurized cuffs, which were placed around the limbs. Thus, the exact detection point of the pulse wave on the artery varied due to the cuff size. They found that all the estimated PWVs in the branches from the heart to the higher and lower limb arteries and the carotid to the femoral artery were significantly increased in the coronary artery disease patients. Often the R-peak of the ECG signal is used for the starting point of the pulse wave propagation from the heart (Naschitz et al 2004). However, the R-peak shows only the electrical activity of the heart, not the starting point of the wave propagation. There is a short period between the heart activation and opening of the valves of the left ventricle, which is called a pre-ejection period. PEP is not constant and depends on the physical condition of the subject. The ejection of the blood to the aorta can be detected from the phonocardiographic or transthoracic impedance signal. The VaSera VS-1500N (Fukuda Denshi, Tokyo, Japan) vascular screening system registers simultaneously ECG, phonocardiographic and pressure wave signals from the limb arteries. However, the PWV values for each arterial branch are not given. Instead, the Cardio-Ankle Vascular Index (CAVI) is calculated, which is based on the PWV estimation. The aortic PWV is of major interest as it is a cardiovascular risk marker of an advanced disease and may predict morbid events in the next five up to ten years (Mancia et al 2007). However, the aortic PWV does not offer any insight into the condition of smaller blood vessels. All the arteries are stiffening and the atherosclerotic disease involves the whole cardiovascular system. The assessment of the peripheral arteries may enable earlier detection of atherosclerosis (Cohn J N 2006). It has been noted that lower limb arteries are particularly altered by atherosclerosis (Laurent et al 2006). However, currently there is no evidence that the PWV measured from the peripheral arteries may predict the cardiovascular event (DeLoach and Townsend 2008). 23

24 3.3. Pulse waveform analysis Pressure wave is generated by the left ventricle contraction and blood ejection into the aorta. The pressure in the aorta increases due to the blood volume increase. The pressure increase in the aorta is dependent on the capability of the arterial wall to stretch and on the blood volume decrease throughout the peripheral beds of the tissue. The blood volume change is connected to the change in the pressure through the arterial compliance. The arterial compliance is decreased due to the increase in the arterial stiffness and in the case of the same stroke volume, a larger pressure wave is generated (Avolio et al 2010). The pulse wave propagation through the vascular tree is a complex process. The nature of wave propagation depends on the elastic properties and geometry of the arteries and blood viscosity. In addition, the pulse wave reflection appears at the location where the artery is branching. The wave reflection and reflected wave magnitude are explained through the wave transmission theory as a mismatch between the impedances of the branching arteries. The impedance is a complex quantity and it is determined by the mechanical properties of the blood and arteries. As a result, the waveform of a propagating pulse wave changes along the arterial tree. Therefore, the pressure waves at the two arterial locations with the same mean pressure can have different pulse pressure and shape (Avolio et al 2010). The pressure waveform at the aortic root depends on the contraction of the left ventricle and the geometric and elastic properties of the arterial system. The distinctive features can be detected from one cardiac cycle long pressure waveform (Figure 2). In the ascending aorta, the peak of the flow velocity wave appears before the peak of the pressure wave, because of the stretch of the elastic artery segment. The inflection point (P i ) on the aortic pressure waveform appears about at the same time with the maximal point of the flow wave. The peak of the flow velocity wave locates at about 30% of the whole ejection duration (Westerhof et al 2006). The augmented component is described through the augmentation index (AIx) and is defined as (Murgo et al 1980): Ps Pi AP AIx, (4) P P PP s d where P s is the systolic blood pressure (maximal value of the pressure waveform), P d is the diastolic blood pressure (minimal value of the pressure waveform), AP is the augmentation pressure and PP is the pulse pressure. The AIx has been found to change through the ageing (Kelly et al 1989). The AIx calculated from the aortic pressure waveform is negative in adolescents and becomes positive for older subjects. The significant changes in the pressure wave morphology are explained through the changes in the structural components of the arterial system, which affects the pulse propagation. In addition, it has been found that the AIx is influenced by the blood pressure and the values are higher in the case of patients with type I diabetes (Wilkinson et al 2000a) and hypercholesterolemia (Wilkinson 24

25 2002), which can be explained through premature vascular ageing (Ravikumar et al 2002, Cockroft and Wilkinson 2002). Figure 2. Averaged radial and aortic pressure waveforms normalized in time with the characteristic points (reproduced according to Avolio et al 2010). The radial pressure waveform includes features similar to those of the aortic waveform registered from the same subject. However, the waveforms are significantly different. The minimal point of the radial artery pressure waveform corresponds to the diastolic pressure similarly to the aortic waveform. The mean pressure of the radial artery waveform is lower from the aortic waveform, but the difference is small and it can be considered equal in the case of healthy subjects. The diastolic and mean pressures measured from the radial or brachial arteries have been used as reference for the aortic waveform calibration. The maximal point in the radial artery waveform corresponds to the systolic blood pressure, but it is situated within the cardiac cycle at a different location (Avolio et al 2010). This causes also the differences between simultaneously measured invasive radial and aortic systolic pressures (Pauca et al 1992). Usually the inflection point (P i ) appears in the radial artery pressure waveform after the first pressure peak, which in this case is the systolic pressure peak. However, due to the ageing P i may become higher than the first peak, forming a systolic pressure peak (Kelly et al 1989). The formed systolic peak in the radial artery waveform is then also connected to the systolic peak in the aortic waveform (Takazawa et al 2007). In some cases the late systolic peak can be visible in the radial artery waveform as a separate peak with the local maximal point or it forms the convexity before the dicrotic notch, as shown in Figure 2. The incisura, which is visible after the systolic peak in the aortic waveform, corresponds to the aortic valve closure. The local minimum after the inflection point on the radial artery waveform is called the dicrotic notch (Avolio et al 2010). The timing of incisura on the aortic waveform correlates with the dicrotic notch in the radial artery waveform and thus the ejection duration can be obtained (Gallagher et al 2004). 25

26 Often the composition of the pulse waveform has been explained through the forward and backward travelling wave. The forward travelling wave is caused by the left ventricle ejection. The backward travelling wave is mainly caused due to the reflection of the forward travelling wave from the branch points of the arteries. It has been reported that the single reflection site is difficult to determine (Westerhof et al 2008). There are many reflection sites in the arterial system and the reflected waves can summate and act as one wave arising from a single reflection point (Nichols and O'Rourke 2005). According to the theory, the stiffness of the arteries is only one variable that determines the AIx. In addition, the aortic pressure waveform depends on the wave characteristics of the left ventricle ejection profile, magnitude and timing of the reflected waves (Kingwell and Gatzka 2002, Nichols and O'Rourke 2005). The pulse waveform analysis allows the detection of changes in the waveforms and attempts have been made to interpret it through the changes in the mechanical properties of the arteries (Hamilton et al 2007). The increase in the aortic augmentation pressure is related to the increase in the arterial stiffness. Due to the higher arterial stiffness the reflected wave propagates with higher velocity and arrives faster back to the aorta from the reflection sites by augmenting the pressure wave and increasing the central systolic blood pressure. The reflected wave arrives back during the systolic phase by increasing the left ventricle afterload (Chirinos and Segers 2010). The earlier arrival of the reflected wave also causes the decrease in the diastolic pressure, which results in the reduction of the coronary artery perfusion pressure (Nichols and O'Rourke 2005). As the pressure waveform changes while travelling through the arterial system, it is important to record the central pressure waveform in order to define the effects on the left ventricle and coronary artery. The direct recording of the aortic pressure wave is invasive. The aortic pressure waveform can be analyzed non-invasively by using the SphygmoCor system. The pressure wave can be registered from the carotid artery by using applanation tonometry. As the carotid artery situates close to the aorta, the waveforms are similar and no transfer function is needed for the analysis. However, the recording of the pressure wave from the carotid artery needs a highly trained person, as the vessel is not supported. In addition, the recording of the pressure wave is difficult in the case of subjects with obesity and carotid stenosis (Rhee et al 2008). The other widely used technique, which is also applied on the SphygmoCor system, involves the pressure wave registration from the radial artery and the aortic waveform is derived by using a generalized transfer function (Pauca et al 2001). The pressure wave recording from the radial artery is preferred as it is supported by bone and the applied pressure by tonometry is easier to handle. The AIx is determined from the derived aortic waveform and it has been used widely as an indicator of aortic stiffness (Rhee et al 2008). Nevertheless, issues regarding the precision of the generalized transfer function have been raised (Hamilton et al 2007). Transfer functions are generally constructed from the data of healthy subjects. It has been found that this approach may cause errors in the case of type II 26

27 diabetes patients and the AIx is overestimated (Hope et al 2004). In addition, it has been suggested that there is no need to use the transfer function for the aortic AIx estimation. The AIx estimated from the radial artery and the AIx estimated from the derived aortic waveform are highly correlated r=0.96 (Millasseau et al 2003). This approach has been applied in the Arteriograph where the aortic AIx is estimated from the brachial artery pressure waveform without using the generalized transfer function (Horváth et al 2010). It has been found that the aortic AIx depends on the heart rate (Wilkinson et al 2000b). Therefore, the AIx value in the SphygmoCor system has been normalized for the heart rate of 75bpm and noted as 4. Photoplethysmographic signal and waveform analysis PPG is a non-invasive optical technique, which can be used to monitor the circulatory changes in the cardiovascular tissue bed. For the PPG signal recording the light source is placed towards the skin surface and light is emitted to the examined tissue. The light is absorbed and scattered in the tissue. Just a small fraction of emitted light intensity can be detected with the photodetector, which is placed on the skin adjacent to the light source (reflection mode) or opposite the measured volume (transmission mode). In the PPG probes, the LEDs and photodiodes or phototransistors are often used, which makes the technology relatively cheap and reproducible. In PPG probes the LEDs are usually in the red or infrared wavelength range. However, the other wavelength ranges are used as well and it is application dependent (Allen 2007). The PPG signal registered from the finger or forehead generally consists of DC and AC components and noise. The amplitude of the AC component can be ten times smaller than the DC component. The DC component depends on the total blood volume in the observed tissue and it varies slowly in time. The AC component of the PPG signal is synchronous with the cardiac cycle and depends on the blood volume and flow changes, blood vessel wall movement, pulsatile pressure and the orientation of red blood cells (Kamal et al 1989, Lindberg and Öberg 1993, Allen 2007). Though origins of the AC component waveform mechanism of the PPG signal have been studied, the phenomenon is still not fully understood (Allen 2007). Generally, it has been accepted that the AC component of the PPG signal can contain valuable information about the cardiovascular system and it corresponds to the heart generated pulse wave. AC and DC components of the PPG signal are affected by respiration and factors that influence local perfusion, such as vasomotion and sympathetic nervous system activity (Johansson and Öberg 1999, Nitzan et al 1998). PPG technology is widely used in the pulse oximeters, however other clinically important physiological parameters such as the breathing and heart rate, can be monitored as well. In addition, the PPG technology has been used for the PTT measurements in order to estimate the vascular ageing (Allen and Murray 2002). Furthermore, the PTT measurement is widely used for the estimation of beat-to-beat blood pressure without cuff. This technique would enable monitoring the blood 27

28 pressure changes over long periods (e.g. 24 hours) (Zheng et al 2013) and providing clinically important dynamical changes, which are predictors of cardiovascular events (Fagard et al 2008, Dolan et al 2005). The PPG waveform is affected by the changes in the mechanical properties of the arteries as the AC component of the signal corresponds to the pulse wave. Therefore, this simple technique may also have a potential for the identification of premature increase in the stiffness of the arteries and may be of considerable value in the prevention of the cardiovascular disease. In all those applications the AC component carries essential information. Therefore, in the subsequent text the term PPG signal corresponds to the AC component of the PPG signal. Noise in the PPG signal can be originated from different sources (Sukor et al 2011). The power line interference causes the 50Hz or 60Hz noise. The ambient light may cause the blinding of the photodetector or modulate the signal. Often the PPG signals are affected by the motion caused noise, which is related to the movement of the sensor or the limb of the subject, where the signal is obtained. In addition, the pulsating volume in the examined tissue is diminished due to the poor perfusion state and therefore the amplitude of the PPG signal is decreased. Consequently, the SNR of the signal is decreased. Similarly to the pressure wave, the foot of the PPG signal has been detected for the PTT estimation (Nitzan et al 2002). Nevertheless, 50% of the raising front level detection has also been used for the PTT measurement, which has been found comparable with the foot detection and it is more appropriate if there is a high probability for motion artifacts in the signal (Lass et al 2004). Often the time delay between the R-peak of the ECG signal and the raising front of the PPG signal is detected for the PTT estimation (Naschitz et al 2004). In addition, the R-peak of the ECG signal has been used for the wave front detection in order to align the recurrences of the PPG signal for the waveform analysis (Allen and Murray 2003). High correlations between the blood pressure and the PTT have been found in the case of physical exercise (Marcinkevics et al 2009). Furthermore, advanced models are proposed for the blood pressure estimation from the PTT measurement and 24- hour monitoring (Poon and Zhang 2005). In addition, for more accurate blood pressure estimation, the PTT is measured without the PEP. For that purpose the time delay is measured between the opening of the aortic valve and the PPG signal raising front (Solà et al 2011). The opening of the aortic valve is detected from the phonocardiographic or transthoracic impedance signals. In the mentioned applications the raising front of the PPG signal must be able to detect accurately, which can be complicated in the case of noise. Different methods have been used to suppress the noise in the PPG signal. The simple FIR bandpass filter can be used in order to limit the unwanted high and low frequency components. However, the gross movement, coughing and breathing patterns, such as deep gasp or yawn, may cause the noise in the PPG signal, which shares the same frequency components, where the harmonic components of the PPG signal are situated. One possibility is to detect the noisy recurrences of the PPG signal and leave them out from the further analysis (Couceiro et al 2012). The 28

29 other approach is to remove the noise by using different signal processing algorithms. Techniques based on wavelet transform (Lee and Zhang 2003) and independent component analysis (Stetson 2004, Kim and Yoo 2006) have been applied. Though, the studies have shown that these methods fail in certain situations (Foo 2006, Yao and Warren 2005). The motion artifacts can be removed from the PPG signals using the Fourier series analysis on cycle by cycle basis (Reddy et al 2008). However, this approach is computationally complex. In addition, the adaptive filters have been used to suppress the noise caused by motion under an assumption that the expected PPG signal is statistically independent of the artifacts. Additional acceleration sensors have to be used with the PPG probe for that purpose (Foo and Wilson 2006, Wood and Asada 2007, Comtois et al 2007). The PPG signal has a slowly changing nature and most of the power of the PPG signal is concentrated into the first 10 harmonic components. The PPG waveform has characteristics similar to those of the pressure waveform and the systolic phase and diastolic phase are separated with the dicrotic notch (Chan et al 2007). The front of the PPG signal is the fastest changing part during one cardiac cycle. The waveform analysis of the PPG signal has been carried out since Dillon and Hertzman firstly described the PPG waveform through the crest time and the height of the dicrotic notch (Dillon and Hertzman 1941). It was found that the height of the dicrotic notch from the baseline depends on the vasoconstriction and vasodilatation effects. In addition, there is an increase in the crest time and disappearance of the dicrotic notch in the subjects with hypertension and arteriosclerosis. The PPG signal waveform depends on the location where the signal is obtained (Allen and Murray 2003). The recorded PPG signal waveform from the finger is almost identical to the pressure waveform recorded from the radial artery. The relationship can be represented by a single transfer function (Millasseau et al 2000). Thus, the mechanical properties of the arteries that are determining the waveform of the radial artery are similarly affecting the waveform of the finger PPG signal. The cardiovascular ageing has a noticeable influence on the PPG signal waveform through the mechanical changes in the arterial system. This change is visible in the PPG waveforms, which are recorded from different locations of the body (Allen and Murray 2003). The increase in the age of the subject and the consequent vascular ageing results in the triangulation of the curvy finger PPG waveform (Hlimonenko et al 2003). These changes are similar to the radial artery (Kelly et al 1989). In the recent studies the PPG waveform has been decomposed by using the Gaussian or log-normal functions in order to separate direct and reflected waves for the cardiovascular system assessment (Rubins 2008, Huotari et al 2009). In addition, the PPG waveform has been analyzed on the frequency domain and the decrease of power in the harmonics is indicated due to the increase in age. This has been suggested as the useful non-invasive measure of ageing and vascular disease (Sherebrin and Sherebrin 1990). A number of parameters can be used to determine the waveform of the pulse wave on the time domain, which can be applied as well for the PPG waveform 29

30 (Korpas et al 2009). On the time domain, the proportions of the pulse wave can be measured, such as the crest time of the pulse wave front, time from the foot of the pulse wave until the maximal point of the diastolic wave, total pulse duration, systolic and diastolic peak amplitude. From the previously mentioned proportions, parameters related to arterial stiffness can be calculated (Millasseau et al 2002, Millasseau et al 2006). In the systolic phase of the PPG waveform similarly to the radial artery pressure waveform, initial wave peak, inflection point, late systolic peak, and dicrotic notch can be seen. Through triangulation of the waveform the distinctive points are difficult to distinguish. The derivatives of the PPG signal can be used for the detection of the mentioned characteristic points on the waveform. The SDPPG signal analysis was first introduced by the Takazawa group to evaluate ageing and increase of arterial stiffness (Takazawa et al 1998). However, the detection of the distinctive points from the PPG waveform is out of scope of this analysis. Instead, the five distinctive wave peaks a, b, c, d, and e from the SDPPG signal are detected and amplitudes of each wave are estimated. The amplitudes are normalized with wave a as follows: b/a, c/a, d/a, e/a. It was found that with the increase of age, the value of b/a (r=0.75; P<0.001) increased and c/a (r=-0.67; P<0.001), d/a (r=-0.72; P<0.001), and e/a (r=-0.25; P<0.001) decreased. According to the results, the ageing index was proposed as follows (Takazawa et al 1998): AGI b c d e. (5) a It was found that the average AGI was elevated in case the subject had any of the following diseases: diabetes mellitus, hypertension, ischemic heart disease, and hypercholesterolemia. The average AGI for 474 healthy subjects was -0.22±0.36 and 126 subjects with the history of a disease had an average AGI -0.06±0.41. It was followed by the studies where the correlation relationship between the SDPPG normalized amplitudes and other physiological parameters was analyzed in more detail. The changes in the amplitudes of the SDPPG waves were analyzed in children and young people by Iketani et al (2000). They found that ageing decreased the b/a ratio and the AGI and increased the c/a and e/a ratios. Overall, the AGI decreased with age between 3 and 18 years and then increased, forming a parabolic relationship. In the study by Bortolotto, the SDPPG amplitude ratios of the finger were compared with the Complior measured carotid-femoral PWV regarding the influencing factors of atherosclerosis and age, in a large hypertensive population (Bortolotto et al 2000). It was found that the average values of the PWV, the AIx of the PPG waveform and the AGI were elevated in the case of hypertensives with atherosclerotic alterations. The relationship between the wave amplitudes of the SDPPG and various cardiovascular risk factors among middle-aged men were analyzed by Otsuka et al 30

31 (2007). They found that the b/a ratio increased with age, hypertension, dyslipidemia, impaired fasting glucose/diabetes mellitus, and lack of regular exercise. Similarly, the d/a ratio decreased with age, hypertension, and alcohol intake 6 up to 7 days per week. In addition, from the femoral artery, the registered PPG signal has been used for the regional and local arterial stiffness estimation by the SDPPG analysis (Grabovskis et al 2011). It was found that the b/a ratio correlated (r=0.729; P<0.0001) with Young s modulus of the femoral artery, which is a measure of arterial stiffness. In addition to the SDPPG parameters, the AIx of the invasive aortic pressure and the finger PPG waveform were compared in the study by the Takazawa group (Takazawa et al 1998). The AIx was estimated as a ratio between the amplitudes of the late and early systolic component. They found high correlation between the two measures of the AIx (r=0.86; P<0.001). This shows the possibility to use the finger PPG signal waveform to estimate the changes in the aortic pressure wave. However, in the later studies, the ratio between the amplitudes of the PPG signal and the diastolic wave peak has been used to estimate arterial stiffness and it is called the reflection index. However, this measure differs from the AIx. The PPG signal has to be twice differentiated for the SDPPG analysis. The differentiator works as a high-pass filter and as a result, higher frequency components are amplified (Proakis and Manolakis 2006). The amplified higher frequency components have to be suppressed, because it consists of unwanted noise. It has been noted that in the case of the PPG waveform analysis it is important to apply appropriate filtering and it is principally essential when using the second derivative analysis (Millasseau et al 2003). In the same study, more detailed information about the band limiting filter of the PPG signal has been given, where the Parks-McClellan digital low-pass filter (aka Remez filter) was used with the following parameters: edge-frequency of 10 Hz, transition-band of 2Hz, pass-band ripple of 0.05 db, stop-band attenuation of 100 db. The edge-frequency of the pass band is the highest frequency of interest in the PPG signal. According to Millasseau et al (2003), the parameters of the filter were claimed to be chosen by unpublished observations, where the filter with the given edge-frequency produced no loss of information. However, the suppression of the higher frequency components has to be enough in order to detect the distinct wave peaks from the SDPPG signal. The band limiting filter for the SDPPG analysis, similar to that of Millasseau et al (2003), has commonly not been described in detail including Takazawa et al (1998), Iketani et al (2000), and Otsuka et al (2007). The cut-off frequency of the low-pass filter has been used between 10 up to 10.6 Hz in many studies. In addition, the transition band of the filter, which plays an important role in the processing of the signal, has not been given in most of the studies (Bortolotto et al 2000, Hashimoto et al 2005, Otsuka et al 2006). 31

32 5. Experimental studies 5.1. Experimental measurement complex (Publication I) A measurement complex for synchronous recording of pulse wave signals from different locations of arterial and peripheral sites as well as other physiological signals registered by different devices was developed (Publication I). The complex provides simultaneous estimation of the PWV and pulse waveform parameters related to the arterial stiffness. The experiments were carried out in the North Estonia Medical Centre. The study was conducted in accordance with the Declaration of Helsinki and formally approved by the Tallinn Ethics Committee on Medical Research. Two reference devices, a SphygmoCor and an Arteriograph, were used in the measurement complex for the aortic AIx and PWV estimation. As those devices are widely used in the clinical studies a possibility was open to compare the PWVs from other segments of the arteries and pulse waveform analysis results. The PWVs were measured from the segments of upper and lower limb arteries in addition to the aortic PWV. In publication I the pulse wave signals were recorded from the elbow, wrist, index finger, forehead, neck, earlobe, knee, and the big toe. However, after the pilot studies some of the locations were changed. After corrections the pulse waves were recorded from the index finger, wrist (radial artery), elbow (brachial artery), forehead, knee (popliteal artery), femoral artery, and the big toes. The pulse waves were registered from the index finger, big toe and forehead by using the PPG technique. The PPG signals were obtained from the finger and the big toe using the commercially available Envitec F clip sensors. The Masimo LNOP TF-I reflectance sensor was used to record the forehead PPG signal. The distance between the LED and the photodetector was 7mm. In both sensors the infrared LED was used, which has peak at 880 nm. The PPG sensors were connected to the lab-built modules, which drives the LED and transforms the photoelectric current from the photodiode into the voltage signal. LED was working in the pulsed mode, which enables the noise from ambient light to be cancelled. In addition, the current of the LED can be adjusted manually from the module and be optimized for the recording. Four PPG modules were built by the author of this thesis. It is difficult to obtain the pulse wave signal from the arteries by using the commercially available PPG sensors. The arteries are situated under other types of tissues and the intensity of diffused light from the deeper layers of the tissue can be low. For that purpose, location specific reflectance sensors should be developed. The piezoelectric transducers were used for the pulse wave recording from the radial, brachial, femoral and popliteal artery. Commercially available piezoelectric transducers MP100 (ADInstruments, USA) were used for the registration of mechanical pulsations of the arteries. The piezoelectric transducers were placed above the artery and the additional pressure was applied with elastic bandage. It was ensured that the applied pressure was not affecting the signal quality (Pilt et al 32

33 2010a, Pilt et al 2010b). Two piezoelectric modules with two input channels were built by the author of this thesis in order to amplify the signals and convert them suitable for an analog-to-digital converter. In addition to the pulse waves the following signals were recorded simultaneously: phonocardiographic, ECG, peripheral pressure wave. For the PWV estimation in the arterial segments from the heart to the lower and upper limb arteries, the phonocardiographic signals were recorded using the ADInstruments cardiomicrophone. From the phonocardiographic signal it is possible to detect the opening of the left ventricle aortic valve and to leave out the PEP in the estimation of the PTT. In addition, the ECG signal was recorded simultaneously and used as a reference signal for the signal processing. ECG signal is stable and less sensitive to motion caused noise. In addition, it is easy to detect the beginning of heart contraction from the R-peak of the ECG signal. The ECG signal was recorded with a commercially available ADInstruments PowerLab 4/20T device in order to ensure the safety of the patient. In addition, the device enables to recording of the phonocardiographic signal. The ECG and phonocardiographic signals were digitized with the sampling rate of 1 khz. Beat-to-beat blood pressure wave was monitored from the right hand index finger by using the Finapres system. The device enables analogue output from where the calibrated pressure wave can be obtained. The outputs of the Finapres system, the PPG and piezoelectric modules were connected to the National Instruments PCI MIO-16-E1 data acquisition card. The card digitizes the analogue signals with a resolution of 12-bit and sampling rate of 1 khz. The synchronization between two analog-to-digital converters was needed. It was carried out by using an additional synchro signal connection between the two devices. The synchronous recording of the signals was tested by the generated sine wave with the frequency of 5 Hz. The signal from the generator output was fed to the inputs of the PowerLab and the data acquisition card. After the analysis it was found that two systems are recording the signals synchronously. In addition, the pilot study was carried out on four subjects in order to test the applicability of the measurement complex. Firstly, the measurements were carried out with reference devices, while the subject was in supine position. Next, all the signals were recorded synchronously with the measurement complex. The signals were processed in MATLAB and PWVs were calculated for the limb arterial segments and compared with the Arteriograph estimated aortic PWV. All the measured PWVs were below 9 m/s. In addition, the PWVs were lower than the aortic PWV in the segments from the elbow to the index finger and from the knee to the big toe. Similar results were also achieved in the earlier study (Hlimonenko et al 2008). In addition, after the pilot studies the ultrasound system was included to the measurement complex in order to obtain the thickness of the arterial wall and the diameter of the vessel lumen. This enables us to estimate the Young s elastic modulus of the artery, which is the measure of stiffness. 33

34 5.2. Noise suppression in photoplethysmographic signal for PTT estimation (Publication II) The registered pulse wave signals can be noisy in case the examined subject has to deal with some kind of physical activity or has unintentional movements during the recordings. Motion caused noise can overlap the frequency components of the pulse wave signal. In publication II the algorithm for the suppression of motion caused noise from the PPG signal was developed and tested on the simulated and recorded PPG signals from the subject. In addition, the averaging effect of the algorithm on the PTT calculation was analyzed. Comb filter has frequency response where the main lobes are regularly spaced over the frequency domain. All the frequency components situated between the main lobes are suppressed. In case the frequency response of the filter is adjusted according to the harmonic components of the desired signal, the noise components between the main lobes are suppressed. In the case of discrete periodic signals x[k], the comb filter is described as follows: y k k x k D x, (6) 2 where y[k] is the output signal of the filter, k is an integer and refers to the sample number in the signal, and D is the coefficient, which determines the frequency response of the filter. The output of the filter is averaging the two consecutive periods in case the filter is adjusted for the periodic signal frequency. The filter can be modified in order to average more than two consecutive periods. The biosignals are recurring but not periodic. Therefore, the comb filter has to be adapted for each recurrence of the PPG signal. In publication II the filter was adapted based on the pulse frequency, which was calculated from the ECG signal. Generally, in MATLAB the developed adaptive comb filter interpolates the r consecutive recurrences of the PPG signal to the same length and calculates the average waveform. The attenuation between the main lobes is increased by enlarging the number of averaged recurrences r. However, using more than four (first side lobe attenuation 11.4dB) or five (first side lobe attenuation 12.1dB) recurrences will not lead to considerable advantage, as with ten recurrences the first side lobe attenuation is 13.1dB. The adaptive comb filters were tested on 24s long simulated PPG and ECG R- peak signals, the frequency of which varied from 1 Hz up to 2 Hz. The white noise was added to the signal. It was found that the adaptive comb filter with r=3, r=4, r=5 attenuated noise 24 db, 32 db, and 39 db, respectively. In addition, the adaptive comb filter was tested on the PPG signals, which were recorded from the forehead. The ECG signal was recorded synchronously. The subject was asked to make squat downs in order to decrease the SNR of the signal. The recorded PPG signal was filtered with high- and low-pass filters with cut-off frequencies of 0.3Hz and 30Hz, respectively. Next, the adaptive comb filter was 34

35 applied. By comparing the frequency spectrums before and after comb filtering it was visible that the frequency components were suppressed between the harmonic components of the PPG signal. Furthermore, the fronts of the PPG signal were able to visually detect after the comb filtering. The adaptive comb filter has an averaging effect on the signal waveform and due to that its influence on the PTT estimation was analyzed by using simulated PPG and ECG R-peak signals. In the first simulated test signal the PTT varied between 0.25 and 0.35 s. The PTT was estimated by using a raw and a comb filtered PPG signal. The comb filter used six recurrences r=6 for the output signal calculation. It was found that the PTTs estimated using the comb filtered signal are delayed about three periods from that of the raw generated PPG signal. Furthermore, in the second simulated test signal the PTT value was changed sharply from 0.35 to 0.25 s. Sixperiod long reaction time was found in case the adaptive comb filter used six recurrences for output signal calculation. In addition, the PTT delay effect was analyzed in the case of the forehead recorded PPG signal. The change in the PTT was obtained with the Valsalva maneuver. The similar PTT delay effect was visible in case the comb filtered PPG signal was used. However, the typical PTT changes of the Valsalva maneuver were mostly visible in case the four recurrences were used for the filter output signal calculation. The developed adaptive comb filter is a simple method for the suppression of the motion caused noise in the PPG signal. However, the output signal of the filter depends on the previous recurrences and due to that the small characteristic changes were reduced. The balance between noise suppression and allowable signal averaging effect was considered in the applications of the filter Second derivative photoplethysmographic signal analysis algorithm (Publication III) The AGI derived from the analysis of the finger SDPPG waveform can be a potential parameter for the estimation of the arterial stiffness and the evaluation of cardiovascular ageing. In publication III the SDPPG analysis algorithm was improved in order to achieve the minimal standard deviation of the AGI. Lower standard deviation of the AGI gives more effective differentiation between the levels of cardiovascular ageing. It is shown in publication III that in the case of a healthy subject the AGI value may differ between the consecutive recurrences by using the SDPPG analysis algorithm (Millasseau et al 2003). Although the subject was in the supine position and it was assumed that the cardiovascular system was not changing during short periods of time, the difference between the minimal and maximal values of the AGI constituted about 39% from the whole scale of the AGI. It was visible that the detected peaks of the SDPPG waveform were situated in the two consecutive recurrences of the PPG signal at the different phases of the cardiac cycle. The different detection of wave peaks in the consecutive recurrences is due to the 35

36 insufficient suppression of noise and higher harmonic components of the PPG signal. The differentiator works as a high-pass filter and it amplifies the noise located at higher frequency components. It has to be suppressed with a low-pass filter. The low-pass filter with a static cut-off frequency suppresses the frequency components of two consecutive recurrences differently, as the frequency components of the PPG signal are dependent on the heart rate. The new SDPPG signal analysis algorithm limits equally the number of harmonic components in each of the recurrence. The analysis algorithm was implemented in MATLAB. For that purpose, the raw SDPPG signal was resampled to ensure that one of the selected recurrence lengths is 1 s long. Next, the Parks-McClellan low-pass filter with a 1 Hz wide transmission band was applied. The copy of the selected recurrence from the filtered SDPPG signal was placed to the waveform matrix. Then the next recurrence was selected from the raw SDPPG signal and the resampling, filtering and copying process was carried out. The selected recurrence was placed to the next row in the waveform matrix. The described processing loop was carried out for all of the recurrences in the SDPPG signal. In addition, parallel with the SDPPG signal processing, the PPG and the fourth derivative of the PPG signal were processed in the same way. As a result, the three waveform matrices were constructed. All the waveforms in the matrices were aligned according to the 50% level point of the PPG signal raising front. Next, the distinct wave peaks a, b, c, d, and e in the SDPPG waveforms were found between the zero crossing points of the fourth derivative of the PPG signal waveform and the amplitude ratios b/a, c/a, d/a, e/a and AGI were calculated. The optimal edge frequency was found for the Parks-McClellan low-pass filter in order to achieve the lowest standard deviation of the SDPPG wave amplitudes, which in the end minimizes the standard deviation of the AGI. Furthermore, the variation in the placement of detected waves on the time domain has to be minimal throughout all the recurrences for a subject. For that purpose, the 1-minute long PPG signals from the index finger were obtained using the measurement complex described in the publication I. The temperature of the room was controlled and kept at 23±1 degrees Celsius. The PPG signals were recorded from 21 healthy and physically active subjects aged from 21 to 66 years. The subjects were grouped by the age as follows: 20-30, 30-40, 40-50, 50-60, Each age group from 20 to 50 years comprised 5 subjects. The other two groups comprised three subjects. The recorded PPG signals were processed with the new SDPPG algorithm and the edge frequency of the Parks-McClellan filter was changed between 4 to 14 Hz with a step of 1 Hz. For comparison, the same signals were processed with the algorithm by the research group of Millasseau (Millasseau et al 2003). It was found that the minimal standard deviation of the amplitude ratios and minimal variations in the placement of the detected waves on the time domain achieved in the case of the edge frequency of the low pass filter was 6 Hz. This edge frequency has been considered optimal for the finger PPG signals. Compared 36

37 to the previous algorithm, the average standard deviation of the AGI was twice lower, which constitutes 5% of the whole scale of the AGI (Takazawa et al 1998). A relatively high correlation (r=0.91) was found between the age and the AGI, which is in agreement with the previously published results (r=0.80) by the Takazawa group (Takazawa et al 1998). The algorithm was also tested on the signals from the diabetes patient group. The diabetes patients may have an increase in the cardiovascular age and elevation in the arterial stiffness due to the sclerotic processes. The diabetes patient group included 20 subjects. The linear regression model was constructed in order to estimate the relation between the age and the AGI. The regression model was constructed based on the results of healthy subjects. The AGI values found were relatively high for the diabetes patients as compared of those of the healthy subjects. However, some of the diabetes patients had AGI values similar to the healthy subjects. It can be caused by the effective treatment of diabetes mellitus and active lifestyle of the subject, which has stopped the premature stiffening of arteries. It was seen from the Bland- Altman plot that the average difference of the diabetes patient group from the regression line was In addition, a significant difference between the two groups was found (P<0.0005). Based on these results, the improved SDPPG algorithm can be used for the estimation of cardiovascular ageing and the subjects with probable increase in arterial stiffness can be differentiated Forehead photoplethysmographic waveform indices for cardiovascular ageing estimation (Publications IV) In publication IV the forehead PPG signal was analyzed by using the SDPPG analysis algorithm introduced in publication III. The aim was to investigate the cardiovascular ageing effect on the forehead PPG signal and to find SDPPG indices for the differentiation of the subjects with increased arterial stiffness. In this study the forehead PPG signals were recorded from 22 supposedly healthy subjects aged from 21 to 50 years and from two diabetes patients at the age of 33 and 27. As the finger and forehead PPG signal recordings were carried out simultaneously, almost half of the recordings were used from the study described in publication III. In this study the aortic PWV was estimated as a reference for the evaluation of the cardiovascular system. The estimated PWV is related to the stiffness of the aorta through the Moens-Korteweg equation. Elevated PWV indicates the increase in the stiffness of the aorta and it is assumed that the stiffness of blood vessels in the forehead has also increased due to the sclerotic processes. The PWV in the aorta was estimated by the TensioMed Arteriograph. It was found that the diabetes patients had the PWV above 10 m/s, which indicates the increase in the stiffness of the aorta. Furthermore, one healthy subject had the PWV of 15.4 m/s. All the other subjects had the PWV below 9 m/s. However, compared to an earlier study (Ohmori et al 2000), no relationship between the aortic PWV and the age was found. 37

38 In this study the optimal low-pass filter edge frequency was found at 6 Hz, which is similar to the study in publication III. The PPG signals were processed and the SDPPG signal amplitude ratios were estimated. It was found that the extracted amplitude ratios b/a (r=0.60), c/a (r=-0.24), d/a (r=-0.59), and e/a (r=-0.50) are dependent on the age. Furthermore, the changes in the amplitude ratios were in the same direction as the finger signal amplitude ratios (Takazawa et al 1998). This means that the changes similar to those of the finger waveform can be seen in the forehead signal, although the waveforms are not identical. In addition, the absolute values of the amplitude ratios from the forehead and the finger signal are different due to the waveform differences. The values of amplitude ratios b/a and d/a form the diabetes patients and from one healthy subject with elevated PWV were noticeably different from healthy subjects. At the amplitude ratio b/a, the values were elevated and at the amplitude ratio d/a, the values were lowered. The changes in the forehead PPG waveform caused by the cardiovascular ageing can be characterized using the SDPPG amplitude ratios. Based on the results it can be assumed that the premature increase in cardiovascular ageing is more evident at the amplitude ratios of b/a and d/a Finger photoplethysmographic waveform index for discrimination of subjects with higher arterial stiffness (Publications V) In publication V the finger PPG waveform index was compared with the SphygmoCor derived aortic AIx to show that the PPGAI can be used for the detection of premature cardiovascular ageing. The PPGAI was obtained using the SDPPG analysis algorithm described in publication III. The amplitudes of the PPG waveform were obtained at the locations of wave peaks b and d. The named amplitudes were normalized with the amplitude of the PPG waveform and denoted as PPGb and PPGd. The PPGAI was calculated as follows: PPGd PPGAI. (7) PPGb In this study the subject groups and participated subjects were the same as in the study of publication III. Furthermore, in this study the recorded signals for publication III were used. During the experiments the pressure waves were obtained from the left hand radial artery with SphygmoCor prior to the PPG signal recordings. The aortic pressure waveform was derived from the radial artery waveform by using the transfer function. The AIx@75 was estimated from the aortic pressure waveform. The strong linear correlation relationship was found between the SphygmoCor derived aortic AIx@75 and the PPGAI (r=0.85). All the data points from healthy subjects and diabetes patients were used in the correlation coefficient calculation. Similarly, a high correlation relation (r=0.86) between the invasive aortic AIx@75 38

39 and the PPG AIx was found by the Takazawa group (Takazawa et al 1998). In addition, the regression model was constructed for the aortic estimation from the PPGAI. The linear relationships between the aortic the PPGAI and the age were investigated by using the data points from healthy subjects. Relatively high correlation relationship was found between the age and the PPGAI (r=0.75) and the regression model was constructed. The data points from healthy subjects were situated close to the regression line. However, the PPGAI values from diabetes patients were noticeably higher from the regression line, which indicates the premature cardiovascular ageing and increase in the arterial stiffness. Similarly, the positive correlation relationship (r=0.58) was found between the age and the aortic The data points were more dispersed around the regression line as compared to the PPGAI. However, the values of the were higher than the regression model line, which is similar to the PPGAI. The data point differences from the regression lines were calculated for the aortic and the PPGAI. For both indices, significant differences between the healthy subject and diabetes patient groups were found. The average difference of diabetes patients from the regression line was 0.26 and the regression model standard deviation was found 0.10 in the case of the PPGAI. Similarly, the average group difference from the regression line was 14.4% and the regression model standard deviation was 12.11%. It was found that the sensitivity, detectability (specificity) and accuracy were equally 75% in the case of PPGAI. The same values for were 55%, 57% and 56%, respectively. Based on the results, the PPGAI provides better discrimination of the subjects with higher arterial stiffness than the aortic It has been shown that the relatively cheap PPG technology and the developed SDPPG analysis algorithm can be applied to discriminate the subjects with premature increase in the arterial stiffness from the healthy subjects. 39

40 Conclusions The results of this study confirm the possibility to detect premature increase in arterial ageing using an inexpensive and non-invasive optical method based on the novel algorithm of the SDPPG waveform analysis. The measurement complex provides synchronous recording of the pulse wave parameters in different sites and other physiological signals required for the investigations. The adaptive comb filter was developed and tested for the motion caused noise suppression to increase the SNR of the PPG signal. The main results of the current study are as follows: The adaptive comb filter attenuated the white noise in the PPG signal by 24 db, 32 db, and 39 db in the case of the filter used for output calculation 3, 4, and 5 recurrences, respectively. In addition, it was possible to increase the PPG signal SNR for the PTT calculation in case the signal was recorded from the forehead and affected by motion caused noise. However, the filter has an averaging effect on the output signal waveform and the PTT calculation. This effect must be considered in the applications of the filter. Compared to the former algorithm, the standard deviation of the AGI was twice lower using the new SDPPG algorithm for the finger PPG waveform analysis. This gives more effective differentiation between the levels of arterial ageing. Furthermore, the diabetes patients had elevated AGI values, which can be explained through increased arterial ageing. A significant difference was found between the groups of healthy subjects and diabetes patients. Changes caused by the arterial ageing in the forehead PPG signal can be characterized using the developed SDPPG signal analysis algorithm. Furthermore, the subjects with increased arterial stiffness had noticeable increase in the SDPPG signal amplitude ratio of b/a and a decrease in the amplitude ratio of d/a, which indicated the premature arterial ageing. In addition, it was found that the changes caused by the arterial ageing in the forehead waveform are of the same direction as the finger waveform. Strong linear relationship was found between the aortic AIx@75 and the PPGAI. The AIx@75 and PPGAI values were elevated in the diabetes patient group. The PPGAI enables better discrimination between healthy subjects and patients with higher arterial stiffness. 40

41 References Allen J 2007 Photoplethysmography and its applications in clinical physiological measurement Physiol Meas 28 R1-R39 Allen J and Murray A 2002 Age-related changes in peripheral pulse timing characteristics at the ears, fingers and toes J Hum Hypertens Allen J and Murray A 2003 Age-related changes in peripheral pulse shape characteristics at various body sites Physiol Meas Alwan A et al 2011 Global status report on noncommunicable disaeses Geneva, World Health Organization Asmar R, Benetos A, Topouchian J, Laurent P, Pannier B, Brisac A M, Target R and Levy B 1995 Assessment of arterial distensibility by automatic pulse wave velocity measurement. Validation and clinical application studies Hypertension Atlas G M 2008 Development and application of a logistic-based systolic model for hemodynamic measurements using the esophageal Doppler monitor Cardiovasc Eng Avolio A P 1980 Multi-branched model of the human arterial system Med Biol Eng Comput Avolio A P, Butlin M and Walsh A 2010 Arterial blood pressure measurement and pulse wave analysis - their role in enhancing cardiovascular assessment Physiol Meas 31 R1-47 Bank A J, Wang H, Holte J E, Mullen K, Shammas R and Kubo S H 1996 Contribution of collagen, elastin, and smooth muscle to in vivo human brachial artery wall stress and elastic modulus Circulation Baulmann J, Schillings U, Rickert S, Uen S, Düsing R, Illyes M, Cziraki A, Nickering G and Mengden T 2008 A new oscillometric method for assessment of arterial stiffness: comparison with tonometric and piezo-electronic methods J Hypertens Bergel D H 1961 The dynamic elastic properties of the arterial wall J Physiol Blacher J, Safar M E, Guerin A P, Pannier B, Marchais S J and London G M 2003 Aortic pulse wave velocity index and mortality in end-stage renal disease Kidney Int Boehmer R D 1987 Continuous, real-time, noninvasive monitor of blood pressure: Penaz methodology applied to the finger J Clin Monit Bortolotto L A, Blacher J, Kondo T, Takazawa K and Safar M E 2000 Assessment of vascular aging and atherosclerosis in hypertensive subjects: second derivative of photoplethysmogram versus pulse wave velocity Am J Hypertens Boutouyrie P, Laurent S, Benetos A, Girerd X J, Hoeks A P and Safar M E 1992 Opposing effects of ageing on distal and proximal large arteries in hypertensives Journal of Hypertension Supplement 10 S

42 Boutouyrie P, Bussy C, Hayoz D, Hengstler J, Dartois N, Laloux B, Brunner H and Laurent S 2000 Local pulse pressure and regression of arterial wall hypertrophy during long term antihypertensive treatment Circulation Boutouyrie P, Briet M, Collin C, Vermeersch S and Pannier B 2009a Assessment of pulse wave velocity Artery Research Boutouyrie P, Revera M and Parati G 2009b Obtaining arterial stiffness indices from simple arm cuff measurements: the holy grail? J Hypertens Boutouyrie P et al 2010 Determinants of pulse wave velocity in healthy people and in the presence of cardiovascular risk factors: 'establishing normal and reference values' Eur Heart J Bramwell J and Hill A 1922 The velocity of the pulse wave in man Proc R Soc Lond Brooks B A, Molyneaux L M and Yue D K 2001 Augmentation of central arterial pressure in Type 2 diabetes Diabet Med Bussy C, Boutouyrie P, Lacolley P, Challande P and Laurent S 2000 Intrinsic stiffness of the carotid artery wall material in essential hypertensives Hypertension Callaghan F J, Babbs C F, Bourland J D and Geddes L A 1984 The relationship between arterial pulse-wave velocity and pulse frequency at different pressures J Med Eng Technol Cavalcante J L, Lima J A, Redheuil A and Al-Mallah M H 2011 Aortic stiffness: current understanding and future directions Journal of the American College of Cardiology Chan G S H, Middleton P M, Branko G C, Wang L and Lovell N H 2007 Automatic detection of left ventricular ejection time from a finger photoplethysmographic pulse oximetry waveform: comparison with Doppler aortic measurement Physiol Meas Cheng K S, Baker C R, Hamilton G, Hoeks A P and Seifalian A M 2002 Arterial elastic properties and cardiovascular risk/event European Journal of Vascular and Endovascular Surgery Chirinos J A 2012 Arterial stiffness: basic concepts and measurement techniques J Cardiovasc Transl Res Chirinos J A and Segers P 2010 Noninvasive evaluation of left ventricular afterload: Part 1: Pressure and flow measurements and basic principles of wave conduction and reflection Hypertension Ciftçi O, Günday M, Calişkan M, Güllü H, Güven A and Müderrisoğlu H 2013 Light cigarette smoking and vascular function Acta Cardiol Chiu Y C, Arand P W, Schroff S G, Feldman T and Carroll J D 1991 Determination of pulse wave velocities with computerised algorithms American Heart Journal Clowes A W and Berceli S A 2000 Mechanisms of vascular atrophy and fibrous cap disruption Annals of the New York Academy of Sciences Cockcroft J R and Wilkinson I B 2002 Arterial stiffness and pulse contour analysis: an age old concept revisited Clin Sci (Lond)

43 Cohen D L and Townsend R R 2011 Update on pathophysiology and treatment of hypertension in the elderly Current Hypertension Reports Cohn J N 2006 Arterial stiffness, vascular disease, and risk of cardiovascular events Circulation Comtois G, Mendelson Y and Ramuka P A 2007 Comparative evaluation of adaptive noise cancellation algorithms for minimizing motion artifacts in a forehead mounted wearable pulse oximeter In Conf Proc IEEE Eng Med Biol Soc Couceiro R, Carvalho P, Paiva R P, Henriques J and Muehlsteff J 2012 Detection of motion artifacts in photoplethysmographic signals based on time and period domain analysis In Conf Proc IEEE Eng Med Biol Soc Creager M A, Lüscher T F, Cosentino F and Beckman J A 2003 Diabetes and vascular disease: pathophysiology, clinical consequences, and medical therapy: Part I Circulation Crowther M A 2005 Pathogenesis of atherosclerosis Hematology Am Soc Hematol Educ Program Cruickshank K, Riste L, Anderson S G, Wright J S, Dunn G and Gosling R G 2002 Aortic pulse-wave velocity and its relationship to mortality in diabetes and glucose intolerance: an integrated index of vascular function? Circulation Currie K D, Proudfoot N A, Timmons B W and MacDonald M J 2010 Noninvasive measures of vascular health are reliable in preschool-aged children Appl Physiol Nutr Metab Dahlöf B, Devereux R B, Kjeldsen S E et al 2002 Cardiovascular morbidity and mortality in the Losartan Intervention For Endpoint reduction in hypertension study (LIFE): a randomised trial against atenolol Lancet DeLoach S S and Townsend R R 2008 Vascular stiffness: its measurement and significance for epidemiologic and outcome studies Clin J Am Soc Nephrol Dillon J B and Hertzman A B 1941 The form of the volume pulse in the finger pad in health, arteriosclerosis, and hypertension Am Heart J Dobrin P B 1999 Distribution of lamellar deformations: implications for properties of the arterial media Hypertension Dolan E, Stanton A, Thijs L, Hinedi K, Atkins N, McClory S, Den Hond E, McCormack P, Staessen J A and O'Brien E 2005 Superiority of ambulatory over clinic blood pressure measurement in predicting mortality: the Dublin outcome study Hypertension Fagard R H, Celis H, Thijs L, Staessen J A, Clement D L, De Buyzere M L and De Bacquer D A 2008 Daytime and nighttime blood pressure as predictors of death and cause-specific cardiovascular events in hypertension Hypertension Feng J and Khir A W 2010 Determination of wave speed and wave separation in the arteries using diameter and velocity Journal of Biomechanics

44 Foo J Y A 2006 Comparison of wavelet transformation and adaptive filtering in restoring artifact-induced time-related measurement Biomedical signal processing and control Foo J and Wilson S 2006 A computational system to optimise noise rejection in photoplethysmography signals during motion or poor perfusion states Medical and Biological Engineering and Computing Franklin S S 2008 Beyond blood pressure: Arterial stiffness as a new biomarker of cardiovascular disease Journal of the American Society of Hypertension: JASH, Franquet A, Avril S, Le Riche R, Badel P, Schneider F C, Boissier C and Favre J P 2013 Identification of the in vivo elastic properties of common carotid arteries from MRI: a study on subjects with and without atherosclerosis J Mech Behav Biomed Mater Grabovskis A, Marcinkevics Z, Lukstina Z et al 2011 Usability of photoplethysmography method in estimation of conduit artery stiffness In Proc of SPIE-OSA Biomedical Optics X1-7 Gallagher D, Adji A and O Rourke M F 2004 Validation of the transfer function technique for generating central from peripheral upper limb pressure waveform Am J Hypertens Gaszner B, Lenkey Z, Illyés M, Sárszegi Z, Horváth I G, Magyari B, Molnár F, Kónyi A and Cziráki A 2012 Comparison of aortic and carotid arterial stiffness parameters in patients with verified coronary artery disease Clin Cardiol Glagov S, Weisenberg E, Zarins C K, Stankunavicius R and Kolettis G J 1987 Compensatory enlargement of human atherosclerotic coronary arteries N Engl J Med Guyton A C and Hall J E 2006 Textbook of Medical Physiology, 11th edition. Elsevier Saunders Hamilton P K, Lockhart C J, Quinn C E and McVeigh G E 2007 Arterial stiffness: clinical relevance, measurement and treatment Clin Sci (Lond) Hashimoto J, Watabe D, Kimura A, Takahashi H, Ohkubo T, Totsune K and Imai Y 2005 Determinants of the second derivative of the finger photoplethysmogram and brachial-ankle pulse-wave velocity: the Ohasama study Am J Hypertens Hennerici M G 2004 The unstable plaque Cerebrovasc Dis 17 (Suppl 3) Herrington D M, Brown W V, Mosca L et al 2004 Relationship between arterial stiffness and subclinical aortic atherosclerosis Circulation Hirai T, Sasayama S, Kawasaki T and Yagi S 1989 Stiffness of systemic arteries in patients with myocardial infarction. A noninvasive method to predict severity of coronary atherosclerosis Circulation Hlimonenko I, Meigas K and Vahisalu R 2003 Waveform analysis of peripheral pulse wave detected in the fingertip with photoplethysmograph Measure Sci Rev

45 Hlimonenko I, Meigas K, Viigimaa M and Temitski K 2008 Aortic and arterial pulse wave velocity in patients with coronary heart disease of different severity Estonian J Eng Hope S A, Tay D B, Meredith I T and Cameron J D 2004 Use of arterial transfer functions for the derivation of central aortic waveform characteristics in subjects with type 2 diabetes and cardiovascular disease Diabetes Care Horváth I G, Németh A, Lenkey Z, Alessandri N, Tufano F, Kis P, Gaszner B and Cziráki 2010 A Invasive validation of a new oscillometric device (Arteriograph) for measuring augmentation index, central blood pressure and aortic pulse wave velocity J Hypertens Hudson B I, Wendt T, Bucciarelli L G, Rong L L, Naka Y, Yan S F and Schmidt A M 2005 Diabetic vascular disease: it s all the RAGE Antioxid Redox Signal Hughes D J, Babbs C F, Geddes L A and Bourland J D 1979 Measurements of Young's modulus of elasticity of the canine aorta with ultrasound Ultrason Imaging Hurst R T and Lee R W 2003 Increased incidence of coronary atherosclerosis in type 2 diabetes mellitus: mechanisms and management Ann Intern Med Huo Y and Ley K 2001 Adhesion molecules and atherogenesis Acta Physiologica Scandinavica Huotari M, Yliaska N, Lantto V, Määttä K and Kostamovaara J 2009 Aortic and arterial stiffness determination by photoplethysmographic technique Procedia Chemistry Iketani Y, Iketani T, Takazawa K and Murata M 2000 Second derivative of photoplethysmogram in children and young people Jpn Circ J Ikonomidis I, Ntai K, Kadoglou N P et al 2013 The evaluation of pulse wave velocity using Arteriograph and Complior apparatus across multiple cohorts of cardiovascular-related diseases Int J Cardiol Johansson A and Öberg P Å 1999 Estimation of respiratory volumes from the photoplethysmographic signal. Part I: experimental results Med Biol Eng Comput Kamal A A R, Harness J B, Irving G and Mearns A J 1989 Skin photoplethysmography a review Comput Methods Programs Biomed Karamanoglu M 2003 Errors in estimating propagation distances in pulse wave velocity Hypertension 41 E8 Kazanavicius E, Gircys R and Vrubliauskas A 2005 Mathematical methods for determining the foot point of the arterial pulse wave and evaluation of proposed methods Information Technology and Control Kelly R, Hayward C, Avolio A and O Rourke M 1989 Noninvasive determination of age-related changes in the human arterial pulse Circulation Khir A W, O Brien A, Gibbs J S R and Parker K H 2001 Determination of wave speed and wave separation in the arteries Journal of Biomechanics

46 Kim B S and Yoo S K 2006 Motion artifact reduction in photoplethysmography using independent component analysis IEEE Trans Biomed Eng Kingwell B A and Gatzka C D 2002 Arterial stiffness and prediction of cardiovascular risk J Hypertens Korpas D, Hálek J and Dolezal L 2009 Parameters describing the pulse wave Physiol Res Korteweg D J 1878 Ueber die Fortpflanzungsgeschwindigkeit des Schalles in elastischen Röhren Annalen der Physik Lass J, Meigas K, Kattai R, Karai D, Kaik J and Rosmann M 2004 Optical and electrical methods for pulse wave transit time measurement and its correlation with arterial blood pressure Proc Estonian Acad Sci Eng Laurent S, Cockcroft J, Van Bortel L et al 2006 Expert consensus document on arterial stiffness: methodological issues and clinical applications Eur Heart J Lee C M and Zhang Y T 2003 Reduction of motion artifacts from photoplethysmographic recordings using a wavelet denoising approach In Proc IEEE EMBS Asian-Pacific Conf on Biomed Eng Lehmann E D, Hopkins K D, Rawesh A, Joseph R C, Kongola K, Coppack S W, Gosling R G 1998 Relation between number of cardiovascular risk factors/events and noninvasive Doppler ultrasound assessments of aortic compliance Hypertension Li H, Cybulsky M I, Gimbrone M A Jr and Libby P 1993 An atherogenic diet rapidly induces VCAM-1, a cytokine-regulatable mononuclear leukocyte adhesion molecule, in rabbit aortic endothelium Arterioscler Thromb Li J K-J, Melbin J, Riffle R A and Noordergraaf A 1981 Pulse wave propagation Circ Res Li J K-J 2004 Dynamics of the Vascular System (Series on Bioengineering & Biomedical Engineering - Vol. 1). World Scientific Pub Co Inc Libby P 2002a Inflammation in atherosclerosis Nature Libby P 2002b Atherosclerosis: a new view Scientific American Libby P, Ridker P M and Maseri A 2002 Inflammation and Atherosclerosis Circulation Lindberg L-G and Öberg P Å 1993 Optical properties of blood in motion Opt Eng Llauradó G, Ceperuelo-Mallafré V, Vilardell C, Simó R, Freixenet N, Vendrell J and González-Clemente J M 2012 Arterial stiffness is increased in patients with type 1 diabetes without cardiovascular disease: a potential role of low-grade inflammation Diabetes Care Mancia G, De Backer G, Dominiczak A et al ESH-ESC practice guidelines for the management of arterial hypertension: ESH-ESC Task Force on the management of arterial hypertension J. Hypertens Marcinkevics Z, Greve M, Aivars J I, Erts R and Zehtabi A H 2009 Relationship between arterial pressure and pulse wave velocity using photoplethysmography during the post-exercise recovery period Biology

47 Meinders J M and Hoeks A P 2004 Simultaneous assessment of diameter and pressure waveforms in the carotid artery Ultrasound Med Biol Miyaki A, Maeda S, Yoshizawa M, Misono M, Sasai H, Shimojo N, Tanaka K and Ajisaka R 2010 Is pentraxin 3 involved in obesity-induced decrease in arterial distensibility? Journal of Atherosclerosis and Thrombosis Millasseau S C, Guigui F G, Kelly R P, Prasad K, Cockcroft J R, Ritter J M and Chowienczyk P J 2000 Noninvasive assessment of the digital volume pulse. Comparison with the peripheral pressure pulse Hypertension Millasseau S C, Kelly R P, Ritter J M and Chowienczyk P J 2002 Determination of age-related increases in large artery stiffness by digital pulse contour analysis Clin Sci (Lond) Millasseau S C, Patel S J, Redwood S R, Ritter J M and Chowienczyk P J 2003 Pressure wave reflection assessed from the peripheral pulse. Is a transfer function necessary? Hypertension Millasseau S C, Stewart A D, Patel S J, Redwood S R and Chowienczyk P J 2005 Evaluation of carotid-femoral pulse wave velocity: influence of timing algorithm and heart rate Hypertension Millasseau S C, Ritter J M, Takazawa K and Chowienczyk P J 2006 Contour analysis of the photoplethysmographic pulse measured at the finger J Hypertens Moens A I 1878 Die Pulskurve. Leiden Molinari F, Zeng G and Suri J S 2010 A state of the art review on intima-media thickness (IMT) measurement and wall segmentation techniques for carotid ultrasound Computer Methods and Programs in Biomedicine Mulders T A, van den Bogaard B, Bakker A, Trip M D, Stroes E S, van den Born B J and Pinto-Sietsma S J 2012 Arterial stiffness is increased in families with premature coronary artery disease Heart Murgo J P, Westerhof N, Giolma J P and Altobelli S A 1980 Aortic input impedance in normal man: relationship to pressure wave forms Circulation Naghavi M, Libby P, Falk E, et al 2003 From vulnerable plaque to vulnerable patient: a call for new definitions and risk assessment strategies: Part I Circulation Naidu M U R, Reddy B M, Yashmaina S, Patnaik A N and Rani P U 2005 Validity and reproducibility of arterial pulse wave velocity measurement using new device with oscillometric technique: A pilot study Biomed Eng Online 4 49 Nam D-H, Lee W-B, Hong Y-S and Lee S-S 2013 Measurement of spatial pulse wave velocity by using a clip-type pulsimeter equipped with a hall sensor and photoplethysmography Sensors Naschitz J E, Bezobchuk S, Mussafia-Priselac R et al 2004 Pulse transit time by R-wave-gated infrared photoplethysmography: review of the literature and personal experience J Clin Monit Comput Nichols W W and O'Rourke M F 2005 Mcdonald s blood flow in arteries. Theoretical, experimental and clinical principles. Oxford University Press 47

48 Nitzan M, Babchenko A, Khanokh B and Landau D 1998 The variability of the photoplethysmographic signal a potential method for the evaluation of the autonomic nervous system Physiol Meas Nitzan M, Khanokh B and Slovik Y 2002 The difference in pulse transit time to the toe and finger measured by photoplethysmography Physiol Meas Nualnim N, Barnes J N, Tarumi T, Renzi C P and Tanaka H 2011 Comparison of central artery elasticity in swimmers, runners, and the sedentary Am J Cardiol Oguri M, Nakamura T, Tamanuki K, Akita C, Kitaoka C, Saikawa Y and Takahashi M 2013 Subclinical arterial stiffness in young children after Kawasaki disease Cardiol Young Ohmori K, Emura S and Takashima T 2000 Risk factors of atherosclerosis and aortic pulse wave velocity Angiology O'Rourke M F, Staessen J A, Vlachopoulos C, Duprez D and Plante G E 2002 Clinical applications of arterial stiffness; definitions and reference values Am J Hypertens O'Rourke M F and Safar M E 2005 Relationship between aortic stiffening and microvascular disease in brain and kidney: Cause and logic of therapy Hypertension Otsuka T, Kawada T, Katsumata M, Ibuki C and Kusama Y 2007 Independent determinants of second derivative of the finger photoplethysmogram among various cardiovascular risk factors in middle-aged men Hypertens Res Pahkala K, Laitinen TT, Heinonen OJ et al 2013 Association of fitness with vascular intima-media thickness and elasticity in adolescence Pediatrics 132 e77-84 Pannier B, Avolio A P, Hoeks A, Mancia G and Takazawa K 2002 Methods and devices for measuring arterial compliance in humans Am J Hypertens Parati G and De Buyzere M 2010 Evaluating aortic stiffness through an arm cuff oscillometric device: is validation against invasive measurements enough? J Hypertens Pauca A L, Wallenhaupt S L, Kon N D and Tucker W Y 1992 Does radial artery pressure accurately reflect aortic pressure? Chest Pauca A L, O Rourke M F and Kon N D 2001 Prospective evaluation of a method for estimating ascending aortic pressure from the radial artery pressure waveform Hypertension Payne R A, Wilkinson I B, and Webb D J 2010 Arterial stiffness and hypertension: Emerging concepts Hypertension Perk J et al 2012 European Guidelines on cardiovascular disease prevention in clinical practice European Heart Journal Pilt K, Meigas K, Viigimaa M, Kaik J, Kattai R and Karai D 2010a Arterial pulse transit time dependence on applied pressure In: IFMBE Proceedings of the 12th Mediterranean Conference on Medical and Biological Engineering and Computing, Thessaloniki, Greece, May 27-30, Pilt K, Meigas K, Viigimaa M, Kaik J, Kattai R and Karai D 2010b Arterial pulse waveform dependence on applied pressure In: Proceedings of

49 International Biennial Baltic Electronics Conference, Tallinn, Estonia, October 4-6, Poon C C and Zhang Y T 2005 Cuff-less and noninvasive measurements of arterial blood pressure by pulse transit time In Conf Proc IEEE Eng Med Biol Soc Proakis J G and Manolakis D K 2006 Digital signal processing 4th edn. Pearson Prentice Hall Qasem A and Avolio A 2008 Determination of aortic pulse wave velocity from waveform decomposition of the central aortic pressure pulse Hypertension Rajzer M W, Wojciechowska W, Klocek M, Palka I, Brzozowska-Kiszka M and Kawecka-Jaszcz K 2008 Comparison of aortic pulse wave velocity measured by three techniques: Complior, SphygmoCor and Arteriograph J Hypertens Rask-Madsen C and King G L 2007 Mechanisms of Disease: endothelial dysfunction in insulin resistance and diabetes Nat Clin Pract Endocrinol Metab Ravikumar R, Deepa R, Shanthirani C and Mohan V 2002 Comparison of carotid intima-media thickness, arterial stiffness, and brachial artery flow mediated dilatation in diabetic and nondiabetic subjects (The Chennai Urban Population Study [CUPS-9]) Am J Cardiol Reddy K A, George B and Jagadeesh Kumar V 2008 Motion Artifact Reduction and Data Compression of Photoplethysmo-graphic Signals utilizing Cycle by Cycle Fourier Series Analysis In Conf Proc of Instrumentation and Measurement Technology IEEE Reneman R S, Meinders J M and Hoeks A P 2005 Non-invasive ultrasound in arterial wall dynamics in humans: what have we learned and what remains to be solved Eur Heart J Rhee M-Y, Lee H-Y and Park J B 2008 Measurements of Arterial Stiffness: Methodological Aspects Korean Circ J Rubins U 2008 Finger and ear photoplethysmogram waveform analysis by fitting with Gaussians Med Biol Eng Comput Rydén L, Grant P J et al 2013 ESC Guidelines on diabetes, pre-diabetes, and cardiovascular diseases developed in collaboration with the EASD: The Task Force on diabetes, pre-diabetes, and cardiovascular diseases of the European Society of Cardiology (ESC) and developed in collaboration with the European Association for the Study of Diabetes (EASD) Eur Heart J Safar M E, London G M and Plante G E 2004 Arterial stiffness and kidney function Hypertension Salvi P, Magnani E, Valbusa F, Agnoletti D, Alecu C, Joly L and Benetos A 2008 Comparative study of methodologies for pulse wave velocity estimation J Hum Hypertens Sherebrin M H and Sherebrin R Z 1990 Frequency Analysis of the Peripheral Pulse Wave Detected in the Finger with a Photoplethysmograph IEEE Trans Biomed Eng

50 Smith S C Jr, Benjamin E J, Bonow R O, Braun L T, Creager M A et al 2011 AHA/ACCF Secondary Prevention and Risk Reduction Therapy for Patients with Coronary and other Atherosclerotic Vascular Disease: 2011 update: a guideline from the American Heart Association and American College of Cardiology Foundation Circulation Solà J, Chételat O, Sartori C, Allemann Y and Rimoldi S F 2011 Chest pulsewave velocity: a novel approach to assess arterial stiffness IEEE Trans Biomed Eng Solomon C G 2003 Reducing cardiovascular risk in type 2 diabetes N Engl J Med Sowers J R, Epstein M and Frohlich E D 2001 Diabetes, hypertension, and cardiovascular disease: an update Hypertension Stetson P F 2004 Independent component analysis of pulse oximetry signals Conf Proc IEEE Eng Med Biol Soc Strong J P, Malcom G T, Newman W P 3rd and Oalmann M C 1992 Early lesions of atherosclerosis in childhood and youth: natural history and risk factors J Am Coll Nutr 11 51S-54S Stary H C, Blankenhorn D H, Chandler A B et al 1992 A definition of the intima of human arteries and of its atherosclerosis-prone regions. A report from the Committee on Vascular Lesions of the Council on Arteriosclerosis, American Heart Association Circulation Stary H C 2003 Atlas of atherosclerosis: progression and regression, 2nd edition. Parthenon Publishing Sukor J A, Redmond S J and Lovell N H 2011 Signal quality measures for pulse oximetry through waveform morphology analysis Physiol Meas Takazawa K, Tanaka N, Fujita M, Matsuoka O, Saiki T, Aikawa M, Tamura S and Ibukiyama C 1998 Assessment of vasoactive agents and vascular aging by the second derivative of photoplethysmogram waveform Hypertension Takazawa K, Kobayashi H, Shindo N, Tanaka N and Yamashina A 2007 Relationship between radial and central arterial pulse wave and evaluation of central aortic pressure using the radial arterial pulse wave Hypertens Res Temitski K, Lauri J, Pilt K and Meigas K 2012 Assessment of algorithms for detecting an arterial pulse pressure wave equiphase point In: Proceedings of 2012 International Biennial Baltic Electronics Conference, Tallinn, Estonia, October 3-5, Tortora G J and Derrickson B 2011 Principles of anatomy and Physiology, 13th edition. John Wiley & Sons Tuzcu E M, Kapadia S R and Tutar E 2001 High prevalence of coronary atherosclerosis in asymptomatic teenagers and young adults: evidence from intravascular ultrasound Circulation Urbina E M, Kimball T R, Khoury P R, Daniels S R and Dolan L M 2010 Increased arterial stiffness is found in adolescents with obesity or obesity-related type 2 diabetes mellitus J Hypertens

51 van Bortel L M, Balkestein E J, van der Heijden-Spek J J et al 2001 Noninvasive assessment of local arterial pulse pressure: comparison of applanation tonometry and echo-tracking J Hypertens van der Heijden-Spek J J, Staessen J A, Fagard R H, Hoeks A P, Boudier H A and van Bortel L M 2000 Effect of age on brachial artery wall properties differs from the aorta and is gender dependent: a population study Hypertension van Popele N M, Grobbee D E, Bots M L, Asmar R, Topouchian J, Reneman R S, Hoeks A P, van der Kuip D A, Hofman A and Witteman J C 2001 Association between arterial stiffness and atherosclerosis: the Rotterdam Study Stroke Verbeke F, Segers P, Heireman S, Vanholder R, Verdonck and Van Bortel L 2005 Noninvasive assessment of local pulse pressure. Importance of brachial-toradial pressure amplification Hypertension Waxman S, Ishibashi F and Muller J E 2006 Detection and treatment of vulnerable plaques and vulnerable patients: novel approaches to prevention of coronary events. Circulation Westerhof N, Bosman F, De Vries C J and Noordergraaf A 1969 Analog studies of the human systemic arterial tree J Biomech Westerhof B E, Guelen I, Westerhof N, Karemaker J M and Avolio A 2006 Quantification of wave reflection in the human aorta from pressure alone: a proof of principle Hypertension Westerhof B E, van den Wijngaard J P, Murgo J P and Westerhof N 2008 Location of a reflection site is elusive: consequences for the calculation of aortic pulse wave velocity Hypertension Wilkinson I B, Fuchs S A, Jansen I M, Spratt J C, Murray G D, Cockcroft J R and Webb D J 1998 Reproducibility of pulse wave velocity and augmentation index measured by pulse wave analysis J Hypertens Wilkinson I B, MacCallum H, Rooijmans D F, Murray G D, Cockcroft J R, McKnight J A and Webb D J 2000a Increased augmentation index and systolic stress in type 1 diabetes mellitus QJM Wilkinson I B, MacCallum H, Flint L, Cockcroft J R, Newby D E and Webb D J 2000b The influence of heart rate on augmentation index and central arterial pressure in humans J Physiol Wilkinson I B, Prasad K, Hall I R, Thomas A, MacCallum H, Webb D J, Frenneaux M P and Cockcroft J R 2002 Increased central pulse pressure and augmentation index in subjects with hypercholesterolemia J Am Coll Cardiol Williams B 1999 The unique vulnerability of diabetic subjects to hypertensive injury J Hum Hypertens 13 S3-8 Woo C H, Shishido T, McClain C, Lim J H, Li J D, Yang J, Yan C and Abe J 2008 Extracellular signal-regulated kinase 5 SUMOylation antagonizes shear stressinduced antiinflammatory response and endothelial nitric oxide synthase expression in endothelial cells Circulation research

52 Wood L B and Asada H 2007 Low Variance Adaptive Filter for Cancelling Motion Artifact in Wearable Photoplethysmogram Sensor Signals In Conf Proc IEEE Eng Med Biol Soc Woodman R J, Kingwell B A, Beilin L J, Hamilton S E, Dart A M and Watts G F 2005 Assessment of central and peripheral arterial stiffness: studies indicating the need to use a combination of techniques Am J Hypertens Xu J 2003 Do we need a better approach for measuring pulse-wave velocity? Ultrasound Med Biol Yao J and Warren S 2005 A short study to assess the potential of independent component analysis for motion artifact separation in wearable pulse oximeter signals In Conf Proc IEEE Eng Med Biol Soc Yusuf S, Sleight P, Pogue J et al 2000 Effects of an Angiotensin-Converting Enzyme Inhibitor, Ramipril, on Cardiovascular Events in High-Risk Patients N Engl J Med Zhang X, Kinnick RR, Fatemi M and Greenleaf J F 2005 Noninvasive method for estimation of complex elastic modulus of arterial vessels IEEE Trans Ultrason Ferroelectr Freq Control Zheng Y, Yan B P, Zhang Y, Yu C M and Poon C C 2013 Wearable cuff-less PTT-based system for overnight blood pressure monitoring In Conf Proc IEEE Eng Med Biol Soc Zieman S J, Melenovsky V and Kass D A 2005 Mechanisms, pathophysiology, and therapy of arterial stiffness Arteriosclerosis, Thrombosis, and Vascular Biology Zulliger M A, Rachev A, and Stergiopulos N 2004 A constitutive formulation of arterial mechanics including vascular smooth muscle tone American Journal of Physiology - Heart and Circulatory Physiology 287 H

53 Author s publications Pilt K, Meigas K, Lass J, Rosmann M (2007) Signal Processing methods for PPG Module to Increase Signal Quality, In: IFMBE proceedings of 11th Mediterranean Conference on Medical and Biomedical Engineering and Computing, Ljubljana, Slovenia, June 26-30, vol. 16, (DOI: / _111). Pilt K, Meigas K, Lass J, Rosmann M, Kaik J (2007) Analogue step-by-step DC component eliminator for 24-hour PPG signal monitoring, In: Proceedings of the 29th Annual International Conference of the IEEE EMBS, Lyon, France, August 23-26, (DOI: /IEMBS ). Pilt K, Meigas K, Rosmann M, Lass J, Kaik J (2008) An Experimental Study of PPG Probe Efficiency Coefficient Determination on Human Body, In: IFMBE Proceedings of 14th Nordic-Baltic Conference on Biomedical Engineering and Medical Physics, Latvia, Riga, June 16 20, vol. 20, (DOI: / _83). Pilt K, Meigas K, Lass J, Rosmann M, Kaik J (2008) Adaptive impulse correlated filter (AICF) improvement for photoplethysmographic signals, In: Proceedings of the 30th Annual International Conference of the IEEE EMBS, Vancouver, Canada, August 20-25, (DOI: /IEMBS ). Pilt K, Meigas K, Lass J, Rosmann M, Kaik J (2008) Adaptive sum comb filter for PPG signals by using ECG signal as reference, In: Proceedings of 2008 International Biennial Baltic Electronics Conference, Tallinn, Estonia, October 6-8, (DOI: /BEC ). Pilt K, Meigas K, Ferenets R, Kaik J (2009) Adjustment of adaptive sum comb filter for PPG signals, In: Proceedings of the 31st Annual International Conference of the IEEE EMBS, Minneapolis, USA, September 2-26, (DOI: /IEMBS ). Pilt K, Meigas K, Karai D, Kaik J (2009) PPG signal processing for pulse delay computing by using adaptive comb filter, In: IFMBE Proceedings of the 11th International Congress of the Medical Physics and Biomedical Engineering, Munich, Germany, September 7-12, (DOI: / _438). (Publication II) Pilt K, Meigas K, Ferenets R, Kaik J (2010) Photoplethysmographic signal processing using adaptive sum comb filter for pulse 53

54 delay measurement, Estonian Journal of Engineering, 16: (DOI: /eng ). Pilt K, Meigas K, Viigimaa M, Kaik J, Kattai R, Karai D (2010) Arterial pulse transit time dependence on applied pressure, In: IFMBE Proceedings of the 12th Mediterranean Conference on Medical and Biological Engineering and Computing, Thessaloniki, Greece, May 27-30, (DOI: / _102). (Publication I) Pilt K, Meigas K, Viigimaa M, Temitski K, Kaik J (2010) An experimental measurement complex for probable estimation of arterial stiffness, In: Proceedings of the 30th Annual International Conference of the IEEE EMBS, Buenos Aires, Argentina, August 31 September 4, (DOI: /IEMBS ). Pilt K, Meigas K, Viigimaa M, Kaik J, Kattai R, Karai D (2010) Arterial pulse waveform dependence on applied pressure, In: Proceedings of 2010 International Biennial Baltic Electronics Conference, Tallinn, Estonia, October 4-6, (DOI: /BEC ). Pilt K, Meigas K, M. Viigimaa, K. Temitski (2011) Possibility to use finapres signal for augmentation index estimation, In: IFMBE Proceedings of the 15th Nordic-Baltic Conference on Biomedical Engineering and Medical Physics, Aalborg, Denmark, June 14-17, vol. 34, (DOI: / _6). Pilt K, Meigas K, Viigimaa M, Temitski K (2011) Possibility to Use Finapres Signal for the Estimation of Aortic Pulse Wave Velocity, In: IFMBE Proceedings of 5th European Conference of the International Federation for Medical and Biological Engineering, Budapest, Hyngary, September 14-18, vol. 37, (DOI: / _136). (Publication IV) Pilt K, Meigas K, Temitski K, Viigimaa M (2012) Second derivative analysis of forehead photoplethysmographic signal in healthy volunteers and diabetes patients, In: IFMBE Proceedings of World Congress on Medical Physics and Biomedical Engineering, Beijing, China, May 26-31, vol. 39, ( / _109). Temitski K, Lauri J, Pilt K, Meigas K (2012) Assessment of algorithms for detecting an arterial pulse pressure wave equiphase point, In: Proceedings of 2012 International Biennial Baltic Electronics Conference, Tallinn, Estonia, October 3-5, (DOI: /BEC ). 54

55 Pilt K, Meigas K, Temitski K, Viigimaa M (2013) The analysis of finger photoplethysmographic waveform in healthy volunteers and diabetes patients In: IFMBE Proceedings of International Symposium on Biomedical Engineering and Medical Physics, Riga, Latvia, October 10-12, 2012, vol. 38, (DOI: / _14). (Publication III) Pilt K, Ferenets R, Meigas K, Lindberg L-G, Temitski K, Viigimaa M (2013) New photoplethysmographic signal analysis algorithm for arterial stiffness estimation, The Scientific World Journal, vol. 2013, Article ID , 9 pages (DOI: /2013/169035). Pilt K, Meigas K, Temitski K, Viigimaa M (2013) The effect of local cold and warm exposure on index finger photoplethysmographic signal waveform, In: Proceedings of the 35th Annual International Conference of the IEEE EMBS, Osaka, Japan, July 3 7, (DOI: /EMBC ). (Publication V) Pilt K, Meigas K, Ferenets R, Temitski K and Viigimaa M (2013) Photoplethysmographic signal waveform index for detection of increased arterial stiffness, Manuscript submitted 55

56 Kokkuvõte Optilise pulsilaine signaali analüüs arterite varase vananemise määramiseks diabeedihaigetel Südame-veresoonkonna haigused on üheks peamiseks surmapõhjuseks maailmas. Vananedes arteri endoteel kulub ja kahjustub ning põhjustab ateroskleroosi ja naastude teket veresoone seinale. Erinevate haiguste korral, seal hulgas diabeedi puhul, on arterite vananemise protsess kiirenenud. Arterite lupjumise tulemusena suureneb nende jäikus, mis põhjustab südame koormuse tõusu, sest veresoonte poolt avaldatud suurema takistuse tõttu on verd raskem läbi soonte pumbata. See põhjustab omakorda erinevaid veresoonkonnaga seotud haiguseid nagu näiteks hüpertensiooni, südame isheemiatõve jne. Veresoonkonna seisundi muutuste varajase avastamise ning selle tulemusena rakendatud õige ravi korral on võimalik edasine haiguse süvenemine ära hoida. Arterite jäikuse hindamiseks töötatakse välja ning võetakse kasutusele üha enam mitteinvasiivseid meetodeid. Enamus meetodeid on suhteliselt keerukad ja kallid ning protseduuride läbiviimiseks on vaja selleks spetsiaalse väljaõppe saanud operaatorit. Arterite seisundi hindamiseks oleks vaja meetodit, mis oleks mitteinvasiivne, odav ning lihtsalt teostatav. Uute ja mitteinvasiivsete meetodite esiletõusuga on hakatud aina tihedamini kasutama pulsilaine levikiirust ja pulsilaine kuju analüüsi. Antud töö eesmärgiks oli uurida optilise meetodi abil salvestatud pulsilaine kuju ja leviaja analüüsi võimalusi varajase arterite jäikuse suurenemise ning sellest tuleneva veresoonkonna vanuse määramiseks. Selleks koostati uurimiskompleks pulsilainete ja teiste füsioloogiliste signaalide mitteinvasiivseks sünkroonseks salvestamiseks. Signaalide salvestused ning muud vajalikud protseduurid viidi läbi Põhja-Eesti Regionaalhaiglas tervetel ning diabeedihaigetel. Töö käigus uuriti ning koostati algoritm liikumisest tingitud mürade vähendamiseks salvestatud optilises pulsilaine signaalis. Arterite jäikuse erinevuste uurimiseks töötati välja optilise pulsilaine signaali kuju analüüsi algoritm, mis põhineb teise tuletise meetodil. Lisaks võrreldi väljatöötatud meetodi ning algoritmi efektiivsust kliinilises praktikas kasutatavate meetoditega, mis võimaldaks eristada normaalse ning suurenenud arterite jäikusega uuritavaid. Töö tulemusena leiti järgmist: Koostatud uurimiskompleks võimaldab sünkroonselt salvestada pulsilaine ja teisi füsioloogilisi signaale. Koostatud algoritm eemaldab optilisest signaalist liikumisest tingitud müra, mille tulemusena on võimalik pulsilaine leviaega edukalt määrata. Väljatöötatud optilise pulsilaine signaali kuju analüüsi algoritm võimaldab eristada arterite jäikust ning seeläbi veresoonkonna vanust tervete ja diabeedihaigete grupi vahel. 56

57 Abstract Optical pulse wave signal analysis for determination of early arterial ageing in diabetic patients Cardiovascular diseases are one of the leading causes of death in the world. Through ageing the endothelium of the artery abrase and get damaged, which causes the atherosclerosis and generation of plaques to the vessel wall. Due to different diseases and diabetes mellitus, the ageing process of the arteries is accelerated. The stiffness of the arteries is increased as a result of calcification, which causes the increase in the load of the heart, because it is more difficult to pump the blood throught the vessels with increased resistance. This in turn causes different diseases connected to the cardiovascular system, for example, hypertension, ischemia etc. Further propagation of the disease can be prevented with premature detection of the changes in the cardiovascular system condition and with applied correct treatment. Different non-invasive methods have been developed lately and applied to estimate the arterial stiffness. However, most of the methods are relatively complex and expensive, furthermore the trained operator is needed for the measurements. There is a need for methods that are non-invasive, inexpensive and easy to perform. The PWV and pulse wave analysis methods have been used more often in the novel non-invasive devices for the arterial stiffness estimation. The aim of the current work was to investigate the possibilities to detect the premature increase in the arterial stiffness and the consequent increase in the cardiovascular ageing using the pulse waveform and the transit time analysis on an optically recorded signal. The measurement complex was built for the non-invasive syncrhronous recording of the pulse wave and other physiological signals. The recording of the signals and other necessary procedures were carried out on healthy subjects and diabetes patients in the North Estonia Medical Centre. During the study the algorithm was investigated and developed for the supression of the motion caused noise in the recorded optical signal. The optical pulse waveform analysis algorithm was developed, based on the second derivative method. In addition, the effectiveness of the developed method was compared with recognized methods used in clinical practice to differentiate the subjects with normal and higher arterial stiffness. This thesis research has resulted in the following findings: The measurement complex compiled enables synchronous recording of the pulse wave and other physiological signals. Constructed algorithm suppresses the motion caused noise from the optical signal and as a result it is possible to estimate successfully the pulse transit time. 57

58 The algorithm developed for the analysis of the optical pulse wave signal enables differentiation of the arterial stiffness and the consequent cardiovascular system age between the healthy and diabetes patient groups. 58

59 APPENDIX 1 PUBLICATIONS Publication I Pilt K, Meigas K, Viigimaa M, Temitski K, Kaik J (2010) An experimental measurement complex for probable estimation of arterial stiffness, In: Proceedings of 30th Annual International Conference of the IEEE EMBS, Buenos Aires, Argentina, August 31 September 4, (DOI: /IEMBS ) IEEE. Reprinted, with permission, from Pilt K, Meigas K, Viigimaa M, Temitski K, Kaik J, An experimental measurement complex for probable estimation of arterial stiffness, Proceedings of IEEE EMBS,

60

61 )ÿr%ÿ&ÿ%ÿÿs!!ÿrÿÿ )ÿ(ÿ Sÿ ÿ zv ÿy Œÿy v~ v{œÿyz z ~ÿtžÿ ÿš ~ ÿ Œÿu Ž ÿ ÿÿÿÿ tuvwÿyvvz{ ÿ}v~ v{~ v{ ÿƒ v v ÿ ÿ~ ÿ} ÿ ˆ Š ÿÿ ÿ ÿÿÿÿÿÿÿÿÿÿ d8=:5[?hÿo=<5bÿd?<[dÿ\7=@?:bÿ\?8@d:ÿq=@=e??bÿd8=:5=h?ÿ/7e=5:l=ÿ?hgÿlm8=ÿd?=lÿ ÿÿ"ÿ ÿ%ÿÿ!ÿÿ ÿ ( ÿÿÿÿÿÿ"'ÿ ÿÿ#ÿ&ÿ%ÿÿÿÿ ÿÿ! ÿÿ!ÿÿÿÿ ÿ%ÿÿÿÿ ÿ)ÿ*#ÿ$ÿÿÿ #ÿ$ÿ!ÿ%ÿ ÿÿÿ%ÿÿÿ ÿÿÿ"ÿ" ÿÿÿÿÿ ÿ=hm7:5=@?57ÿ567ÿ;6?h@7:ÿ=hÿ>?8?e7578:ÿ9fÿ>d<:7ÿc?m7f98ebÿ C6=;6ÿ?87ÿ;?D:7GÿJAÿ567ÿ?8578=?<ÿ:5=FH7:,ÿ+Hÿ?GG=5=9Hÿ567ÿ 95678ÿ>6A:=9<9@=;?<ÿ:=@H?<:ÿ?HGÿ>?8?E7578:ÿ?87ÿE7?:D87Gÿ F89Eÿ:DJ[7;5BÿC6=;6ÿG7:;8=J7:ÿ567ÿ;?8G=9M?:;D<?8ÿ:A:57EBÿ ÿÿÿ"!#ÿ =:ÿg=?@h9:7gÿ=hÿ7?8<aÿ:5?g=debÿ567ÿ>?569@7h7:=:ÿ;?hÿj7ÿ :59>>7GÿC=56ÿ8=@65ÿ587?5E7H5,ÿK7F987ÿ567ÿF?5:ÿ?87ÿJD=<GD>ÿ9Hÿ 56789:;<789:=:ÿ=:ÿ>?569<9@ABÿC6=;6ÿ;?D:7:ÿ567ÿE9:5ÿ9Fÿ 567ÿ7?8<AÿG7?56:ÿ=HÿID89>7,ÿ+Hÿ;?:7ÿ567ÿ?56789:;<789:=:ÿ +,ÿ+./01234/+1.ÿ ÿÿ ÿÿÿ:d;6ÿ?:ÿ7<7;589;?8g=9@8?>6=;ÿni4nrÿ:=@h?<bÿj<99gÿ>87:d87bÿ 75;,ÿ/67ÿ7j>78=E7H5:ÿ?87ÿ><?HH7Gÿ59ÿ;?8Aÿ9D5ÿ9Hÿ<?8@78ÿ HDEJ78ÿ9Fÿ>?5=7H5:ÿC=56ÿ;989H?8AÿG=:7?:7ÿ?HGÿG=?J757:,ÿ ÿ_98ÿ567ÿ:ah;689h9d:ÿe7?:d87e7h5ÿ9fÿ567ÿ>6a:=9<9@=;?<ÿ :=@H?<:BÿC6=;6ÿ?87ÿ98=@=H?57GÿF89EÿG=F787H5ÿG7M=;7:Bÿ567ÿ E7?:D87E7H5ÿ;9E><7jÿ=:ÿH77G7GÿJD=<5,ÿS:ÿ567ÿ7j>78=E7H5:ÿ ÿ.de789d:ÿe7569g:ÿ?hgÿg7m=;7:ÿ6?m7ÿj77hÿg7m7<9>7gÿ59ÿ C?<:ÿ9Fÿ?8578ABÿ567ÿF=8:5ÿ;6?H@7:ÿ?87ÿ5?L=H@ÿ><?;7ÿ=Hÿ567ÿC?<:ÿ?87ÿ;?8=7Gÿ9D5ÿ=Hÿ69:>=5?<Bÿ567ÿ:A:57EÿH77G:ÿ59ÿJ7ÿE9J=<7Bÿ 9FÿJ<99GÿM7:7<:,ÿ/67ÿC?<:ÿJ7;9E7ÿ:5=F78ÿ?HGÿ56=;L78,ÿ+Hÿ ;9E>?;5ÿ?HGÿ7?:Aÿ59ÿD:7,ÿ/67ÿ5A>7:ÿ9Fÿ:7H:98:ÿ:69D<GÿJ7ÿ 56=:ÿ:5?G=DEÿ567ÿ>?569@7H7:=:ÿ9Fÿ?56789:;<789:=:ÿ;?HÿJ7ÿ :7<7;57Gÿ?;;98G=H@ÿ59ÿJ9GAÿ<9;?5=9HBÿC6787ÿ567ÿ>D<:7ÿ :59>>7G,ÿ C?M7F98Eÿ=:ÿ87@=:5787G,ÿ+,ÿ\I/o12Zÿ?:7:ÿ567ÿ:5=FH7:ÿ9Fÿ567ÿ?8578=7:Bÿ=H;<DG=H@ÿ>D<:7ÿC?M7ÿ M7<9;=5AÿNOPQRÿE7?:D87E7H5:ÿ?HGÿ>D<:7ÿC?M7F98Eÿ?H?<A:=:,ÿ C6=;6ÿ;?HÿJ7ÿD:7GÿF98ÿ567ÿ7?8<AÿG=?@H9:=:ÿ9Fÿ?56789:;<789:=:,ÿ S8578=?<ÿ:5=FH7:ÿ8=:7ÿ;?D:7:ÿ567ÿOPQÿ=H;87?:7ÿTUVWX,ÿ/67ÿ 87;98G7GÿC=56ÿE7?:D87E7H5ÿ;9E><7jÿF89Eÿ:DJ[7;5ÿN_=@D87ÿUR,ÿ +Hÿ9D8ÿ>87M=9D:ÿ87:7?8;6ÿ56787ÿ6?M7ÿJ77Hÿ:5DG=7Gÿ567ÿ :5=FH7:ÿ9Fÿ?8578=7:ÿ;?HÿJ7ÿ7:5=E?57Gÿ7?:=<Aÿ?HGÿH9HV OD<:7ÿC?M7ÿ:=@H?<:ÿ?87ÿ87@=:5787GÿF89Eÿ=HG7jÿF=H@78BÿC8=:5Bÿ 87<?5=9H:6=>:ÿJ75C77HÿOPQÿ?HGÿ567ÿG7>56ÿ9Fÿ?56789:;<789:=:ÿ =HM?:=M7<AÿC=56ÿ87;9@H=Y7GÿG7M=;7:ÿ?:ÿZ>6A@E9;98ÿ?HGÿ 7<J9CBÿH7;LBÿ7?8<9J7Bÿ57E><7ÿ9Fÿ567ÿ67?GBÿLH77ÿ?HGÿ567ÿJ=@ÿ 2D8=H@ÿ567ÿ7j>78=E7H5ÿG=F787H5ÿ>6A:=9<9@=;?<ÿ:=@H?<:ÿ?87ÿ 9Hÿ>?5=7H5:ÿC=56ÿ567ÿ;989H?8Aÿ67?85ÿG=:7?:7,ÿ+5ÿ:69C7Gÿ56?5ÿ S8578=9@8?>6,ÿP=56ÿJ956ÿG7M=;7:ÿ=5ÿ=:ÿ>9:=J<7ÿ59ÿ7:5=E?57ÿ567ÿ >?5=7H5:ÿC=56ÿ;989H?8Aÿ67?85ÿG=:7?:7ÿ6?Gÿ=H;87?:7GÿOPQÿ=Hÿ OPQÿ=Hÿ?985?,ÿ?985?Bÿ<7@ÿ?HGÿ?8Eÿ?8578=7:ÿ;9E>?87GÿC=56ÿ67?<56Aÿ:DJ[7;5:,ÿ 1D8ÿ:5DGAÿ?=Eÿ=:ÿ59ÿ?H?<AY7ÿ567ÿ>6A:=9<9@=;?<ÿ>?8?E7578:Bÿ 67?Gÿ?HGÿH7;Lÿ?87ÿ87@=:5787GÿJAÿD:=H@ÿ>6959V 597,ÿIHDE78?57Gÿ>D<:7ÿC?M7ÿ:=@H?<ÿ87@=:58?5=9Hÿ<9;?5=9H:ÿ?87ÿ ><756A:E9@8?>6=;ÿNOOnRÿE7569G,ÿOOnÿ=:ÿ?ÿH9HV=HM?:=M7ÿ 9Hÿ?8578=7:ÿ?HGÿ?<:9ÿ=Hÿ>78=>678A,ÿ+Hÿ?GG=5=9Hÿ59ÿ>D<:7ÿC?M7ÿ 9>5=;?<ÿ57;6H=pD7ÿF98ÿE7?:D8=H@ÿ;6?H@7:ÿ=Hÿ?ÿJ<99Gÿ :=@H?<:ÿ567ÿ7<7;589;?8G=9@8?>6=;Bÿ>69H9;?8G=9@8?>6=;ÿ?HGÿ ÿ+hÿf9<9c=h@ÿc=g78ÿ87:7?8;6ÿ567ÿopqÿ=:ÿ><?hh7gÿ59ÿ ;=8;D<?5=9Hÿ?HGÿ6?:ÿJ77HÿD:7GÿE?=H<AÿF98ÿE9H=598=H@ÿJ<99Gÿ >78=>678?<ÿJ<99Gÿ>87:D87ÿ:=@H?<:ÿ?87ÿ87@=:5787Gÿ E7?:D87ÿF89EÿG=F787H5ÿ?8578=7:ÿ:=ED<5?H79D:<Aÿ?HGÿ59ÿ>78FD:=9Hÿ=Hÿ:L=HÿT]X,ÿ/67ÿ>D<:?5=H@ÿS4ÿ;9E>9H7H5ÿ9Fÿ567ÿ :AH;689H9D:<A,ÿ 87@=:5787GÿOOnÿ:=@H?<ÿ;987:>9HG:ÿ59ÿ>D<:7ÿC?M7,ÿ /67ÿ>D<:7ÿC?M7:ÿF89Eÿ=HG7jÿF=H@78Bÿ7?8<9J7Bÿ57E><7ÿ9Fÿ567ÿ 87F<7;5=9Hÿ?HGÿ567ÿ58?H:E=:=9HÿE9G7,ÿ+Hÿ567ÿ87F<7;5=9HÿE9G7Bÿ?ÿ>6959G=9G7ÿ=:ÿ><?;7Gÿ?G[?;7H5ÿ59ÿ567ÿqI2ÿ?HGÿG=87;57Gÿ59V I:59H=?HÿZ;=7H;7ÿ_9DHG?5=9HÿN@8?H5ÿ.9,ÿ`a^bRBÿJAÿ567ÿI:59H=?Hÿ5?8@757Gÿ F=H?H;=H@ÿ>89[7;5ÿZ_^Uc^^W`:^`Bÿ?HGÿJAÿ567ÿID89>7?Hÿ3H=9Hÿ5689D@6ÿ567ÿ ID89>7?Hÿ07@=9H?<ÿ27M7<9>E7H5ÿ_DHG,ÿ 3H=M78:=5Aÿ9Fÿ/7;6H9<9@ABÿ/?<=HHBÿUe^fbÿI:59H=?ÿN>69H7gÿh]`WÿbW^WW^^iÿ ÿ\?hd:;8=>5ÿ87;7=m7gÿs>8=<ÿw]bÿw^u^,ÿ/6=:ÿ:5dgaÿc?:ÿ:d>>9857gÿjaÿ567ÿ C?8G:ÿ567ÿ:L=H,ÿ/67ÿ>6959G=9G7ÿE7?:D87:ÿ567ÿ87F<7;57Gÿ?HGÿ :;?5787Gÿ<=@65ÿ=H57H:=5AÿF89Eÿ567ÿ:L=Hÿ:D8F?;7,ÿ+Hÿ567ÿ58?H:V E=:=9HÿE9G7Bÿ567ÿ>6959G=9G7ÿ?HGÿ567ÿ<=@65ÿ:9D8;7ÿ?87ÿ><?;7Gÿ 9Hÿ9>>9:=57ÿ:=G7:ÿ9Fÿ567ÿE7?:D87GÿM9<DE7,ÿ/67ÿ>6959G=9G7ÿ /6787ÿ?87ÿ5C9ÿE?=HÿOOnÿ:7H:98ÿG7:=@HÿE9G7:gÿ567ÿ 27H=:ÿd?8?=ÿ?87ÿC=56ÿ567ÿ27>?85E7H5ÿ9FÿK=9E7G=;?<ÿIH@=H778=H@Bÿ/?<=HHÿ 3H=M78:=5Aÿ9Fÿ/7;6H9<9@ABÿ/?<=HHBÿUe^fbÿI:59H=?ÿN7VE?=<gÿL?<[Dk;J,5D,77Bÿ E?8@D:k;J,5D,77BÿL8=:5=H?k;J,5D,77Bÿ[A8=k;J,5D,77BÿG7H=:k;J,5D,77R,ÿ F?jgÿh]`WÿbW^WW^Uiÿ7VE?=<gÿL8=:5[?H,>=<5k;J,5D,77R,ÿ d8=:5[?hÿo=<5ÿ=:ÿc=56ÿ567ÿ27>?85e7h5ÿ9fÿk=9e7g=;?<ÿih@=h778=h@bÿ/?<=hhÿ E7?:D87:ÿ567ÿ58?H:E=57Gÿ<=@65ÿ=H57H:=5A,ÿ/67ÿ58?H:E=:=9Hÿ :7H:98ÿE7?:D87E7H5ÿ:=57:ÿ?87ÿ<=E=57GBÿJ7;?D:7ÿ9Fÿ@79E758=;:Bÿ š šš š š œ œž Ÿ d?<[dÿ\7=@?:bÿ\?8@d:ÿq=@=e??bÿd8=:5=h?ÿ/7e=5:l=bÿlm8=ÿd?=lÿ?hgÿ ÿ ÿ 7?8<9J7ÿJAÿD:=H@ÿ;9EE78;=?<Aÿ?M?=<?J<7ÿ58?H:E=:=9Hÿ :7H:98:,ÿ2D7ÿ59ÿ567ÿ;<=>ÿ5A>7ÿ9Fÿ;9H:58D;5=9Hÿ567ÿ:7H:98:ÿ;?Hÿ C6787?:ÿ?ÿ87F<7;5=9HÿE9G7ÿ:7H:98ÿ;?HÿJ7ÿ><?;7Gÿ?5ÿ?HAÿ>9=H5ÿ 9Hÿ567ÿ:L=Hÿ:D8F?;7,ÿ /67ÿOOnÿ:=@H?<ÿ=:ÿ87@=:5787GÿF89Eÿ=HG7jÿF=H@78ÿ?HGÿF89Eÿ

62 ÿ ÿ 12ÿ234567ÿ8592ÿÿ23422ÿ635ÿÿ262ÿ8ÿ2ÿ 23ÿ3ÿ2ÿ2ÿ28625ÿ2ÿ4244ÿ32ÿ42ÿÿ54ÿ 2423ÿ32ÿ122ÿ42ÿ26ÿ3934ÿ286232ÿ 4244ÿ3ÿ26ÿ852ÿ3ÿ23ÿ65ÿ34545ÿ2ÿ 4244ÿ2ÿ25435ÿ8ÿÿ4536ÿ37ÿ12ÿ5856ÿ 5ÿÿ2845ÿ 282ÿ2ÿ4536ÿ25ÿ24ÿ 2ÿ16ÿ5635ÿ3ÿ12ÿ45632ÿ5ÿ42ÿ6354ÿ ÿ45364ÿ32ÿ234567ÿ1352ÿ8ÿ25467ÿ24512ÿ ÿÿ4536ÿ25435ÿ54ÿ5856ÿ8ÿ!54"ÿ 261!"ÿ22"ÿ12342ÿ2ÿ32524ÿ32ÿ52ÿ2ÿ2ÿ72ÿ 8ÿ5424ÿ#64ÿ2ÿ15ÿ2ÿ37ÿ32ÿÿ2845ÿ3ÿ2ÿ 4536ÿ54ÿ3ÿÿ135ÿ!5ÿ5ÿ ÿ35ÿ $%&'ÿÿ42ÿ6354ÿ2ÿ52(2625ÿ2ÿ54ÿ42ÿ )5ÿ52(2625ÿ424ÿ5ÿ54ÿ45162ÿÿ2542ÿ 23536ÿ6435ÿ8ÿ32524ÿÿ54ÿ ÿ4562ÿ 2ÿ8ÿ2ÿ225ÿ8ÿ2ÿ642ÿ!32ÿ 7ÿ635ÿ2ÿ 52(2625ÿ424ÿ312ÿ2ÿ327ÿ3ÿ3675ÿ2ÿ 35536ÿ242ÿÿ5"ÿ2ÿ642ÿ!32ÿ4536ÿ3ÿ12ÿ 1352ÿ*ÿ2ÿ432ÿ52ÿ2ÿ3652ÿ242ÿ3ÿ382ÿ 2ÿ642ÿ!32ÿ2657ÿ3ÿ54ÿ!328ÿ%56ÿ2ÿ2ÿ 5536ÿ242ÿ6226ÿ2ÿ58622ÿ54ÿÿ523162ÿ+,-ÿÿ 54ÿ2423ÿ32ÿ122ÿ42ÿ#.424ÿ/00ÿ 3424ÿ 24524ÿ2ÿ642ÿ!32ÿ45364ÿ2ÿ12ÿ3ÿ 3537ÿ45364ÿ32ÿ25422ÿ47467ÿ 5ÿ2ÿ!62ÿ29252ÿ12ÿ4536ÿ245124ÿ2ÿ ÿ3557ÿ8ÿ2ÿ23ÿÿ355ÿ5ÿ54ÿ2347ÿÿ 2252ÿ2ÿ1255ÿ8ÿ23ÿ35ÿ25ÿ8ÿ2ÿ &23ÿ8ÿ2ÿ4536ÿ#4ÿ2ÿ12ÿ4536ÿ54ÿ2ÿ43162ÿ3ÿ 624ÿ58622ÿ17ÿ224ÿ8ÿ4132"ÿ3ÿ25422ÿ 642ÿ!32ÿ45364ÿ32"ÿ5ÿ3ÿ12ÿ42ÿ34ÿ28222ÿ8ÿ2ÿ 8652ÿ4536ÿ245ÿÿ29362ÿÿ22ÿ2ÿ5ÿ 342ÿ5424ÿ8ÿ2ÿÿ4536"ÿ!5ÿ54ÿ25422ÿ8ÿ 2ÿ+4-ÿ 3535ÿ4536ÿ3325(24ÿ2ÿ23ÿ2632ÿ 2424"ÿ!5ÿ32ÿ ÿÿ345ÿ45364ÿÿ 82527ÿ42ÿ8ÿ60ÿÿÿ7008(ÿ32ÿ4532ÿ2ÿ "ÿ!5ÿ32ÿ24515ÿ2ÿ22ÿ8ÿ35ÿ 362ÿ+9-ÿ 7ÿ45ÿ12ÿ3ÿ3535ÿ45364ÿ5ÿ 54ÿ45162ÿÿ2342ÿ5267ÿ2ÿ22325ÿ25ÿ3ÿ 2325ÿ35ÿ5ÿ2ÿ23ÿ762ÿÿ54ÿ364ÿ45162ÿÿ 2342ÿ2ÿ):4ÿ435ÿ8ÿ25ÿ8ÿ2ÿ35ÿ362ÿ ÿ2342ÿ5ÿ652ÿ2ÿ16ÿ242ÿ324ÿ2ÿ 5324ÿ54ÿ42ÿ34ÿ28222ÿ252ÿÿ54ÿ45162ÿÿ2ÿ2ÿ 24535ÿ8ÿ16ÿ242ÿ5ÿ2ÿ!62ÿ29252ÿ *ÿ2ÿ432ÿ52ÿ5ÿ34ÿÿ12ÿ32ÿ3ÿ3ÿ5324ÿ 524ÿ2ÿ5835ÿ31ÿ16ÿ242ÿ5ÿ2527ÿÿ 3ÿ12ÿ342"ÿ3ÿ45563ÿ324ÿ5ÿ16ÿ242ÿ32ÿ 364ÿ35ÿ632ÿ5ÿ32524"ÿ!562ÿ2ÿ4132ÿ54ÿ675ÿ!ÿ ÿ355ÿÿ5324ÿ2ÿ#253ÿ3ÿ%7ÿ 32ÿ42ÿ34ÿ28222ÿ2524ÿÿ24532ÿ2ÿ32536ÿ 45824ÿ 7ÿ45ÿ%7ÿ2ÿ242ÿ!32ÿ54ÿ 25422ÿ8ÿ3536ÿ327ÿ!5ÿ22ÿ#253ÿ 6424ÿ!5ÿ8ÿ2ÿ16ÿ86!ÿ5ÿ13536ÿ327ÿ3ÿ2ÿ 242ÿ!324ÿÿ5822ÿ2424ÿ32ÿ25422ÿ)5ÿ 1ÿ2524ÿ2ÿ25422ÿ242ÿ!32ÿ432ÿ54ÿ3367(2ÿ 3ÿ2ÿ!ÿ35ÿ33224ÿ32ÿ24532;ÿ3235ÿ 529ÿ3ÿ):ÿ5ÿ33<ÿ!5ÿ3325(2ÿ2ÿ45824ÿ8ÿ 32524ÿ2ÿ234224ÿ!5ÿ%7ÿ3ÿ #253ÿ32ÿ352ÿÿ ÿ8ÿ2236ÿ 474ÿ4536ÿ25ÿ24ÿ *ÿ52ÿ=ÿ2ÿ16ÿ533ÿ8ÿ23422ÿ629ÿ54ÿ 52ÿ2ÿ3!ÿÿ45364ÿ32ÿ25422ÿ!5ÿ631156ÿ 624ÿ2ÿ2ÿ8ÿ2ÿÿ424ÿ>1.ÿ3ÿ25422ÿ ÿ4536ÿ35ÿ3ÿ12ÿ42ÿ3367ÿ8ÿ2ÿ62ÿ54ÿ!37ÿ2ÿÿ62ÿ54ÿ3342ÿ8ÿ2ÿ352ÿ45ÿ 33224ÿ3ÿ2ÿ4536ÿ54ÿ25422ÿ!5ÿ5ÿ%&ÿ %556367ÿ2ÿ3!ÿ52(2625ÿ45364ÿ32ÿ25422ÿ!5ÿ ÿ365852"ÿ!5ÿ34ÿ ÿ35ÿ 2ÿÿ45364ÿ8ÿ624ÿ32ÿ55365(2ÿ!5ÿ ÿ 5ÿ=ÿ ÿ23422ÿ629ÿ16ÿ533ÿ!5ÿ33ÿ 86!ÿ5254ÿ ÿ 5ÿ/ÿ2ÿ6354ÿ8ÿ2ÿ4244ÿÿ3ÿ17ÿ

63 ÿ ÿ ÿ ÿ8ÿ8ÿ232ÿ ÿ2ÿ 43ÿ92746ÿ6ÿ!"ÿ#44727ÿ3ÿ$4629ÿ 26275ÿ533ÿ29ÿ5663ÿ35ÿ232ÿ ÿ2ÿ35ÿ 493ÿ3ÿ427ÿ%755ÿ9ÿ2&"ÿ 'ÿ()ÿ4!5734ÿ26ÿ$4629ÿ946279ÿ2ÿ56435ÿ 46ÿ56746ÿ26ÿ5ÿ35ÿ52)ÿ4ÿ49ÿ436ÿ46ÿ *2%+8,ÿ6&45663"ÿ6ÿ$4ÿ-ÿ49ÿ4&6ÿ52ÿ 53746ÿ5ÿ35ÿ94627ÿ329"ÿ$5ÿ%3ÿ ÿ27ÿ3ÿ ÿ2ÿ4972ÿ56ÿ3ÿ96ÿ46ÿ.463"ÿ/7ÿ 3ÿ946279ÿ2ÿ5ÿ2ÿ4353ÿ946ÿ26ÿ ÿ 473ÿ5ÿ27543"ÿ#347ÿ5ÿ ÿ&ÿ94627ÿ26ÿ%ÿ 473ÿ43ÿ75ÿ26ÿ4ÿ29ÿ4739)ÿ4ÿ35ÿ 6ÿ26ÿ%ÿ93ÿ2627"ÿ86ÿ243456ÿ3ÿ5ÿ746ÿ 4636ÿ26ÿ%ÿ744623ÿ43ÿ01!ÿ653ÿ473"ÿ$5ÿ3ÿ 5ÿ232ÿ& ÿ3ÿ2.427)ÿ46427ÿ26ÿ 2743ÿ5ÿ3ÿ94627ÿ2ÿ4972ÿ56746ÿ5ÿ&ÿ4&ÿ 9569"ÿ&ÿ32ÿ5ÿ3ÿ94627ÿ29ÿ2743ÿ464235)ÿ 4ÿ369ÿ4635ÿÿ46ÿ29ÿ3ÿ94627ÿ49ÿ53ÿ5ÿ26)ÿ43ÿ 355ÿ75ÿ5ÿ355ÿ4ÿ2743"ÿ/9ÿ3ÿ2ÿÿ946279ÿ35ÿ 56435ÿ3ÿ464235ÿ79ÿ5235ÿ35ÿ3ÿ293ÿ5&&4ÿ3ÿ 2743ÿ5ÿ3ÿ946279"ÿ 'ÿ2ÿ(ÿ26ÿ ÿ946279ÿ2ÿ 5%3246ÿ26ÿ443274!ÿ43ÿ/ ÿ5*2%ÿ 3451'ÿ&4"ÿ'49ÿ&4ÿ29ÿ53427ÿ495723ÿ(ÿ 2744)ÿ4ÿ699ÿ3ÿ23463ÿ923"ÿ/9ÿ3ÿ5*2%ÿ 3451'ÿ29ÿ232ÿ ÿ567ÿ35ÿ6#7ÿ53ÿ36ÿ 3ÿ946279ÿ2ÿ5ÿ43ÿ23ÿ#532"ÿ'ÿ946279ÿ2ÿ 4434!ÿ944727ÿ35ÿ232ÿ ÿ2ÿ43ÿ92746ÿ 6ÿ!"ÿ /9ÿ3ÿ(ÿ26ÿ ÿ946279ÿ2ÿ5ÿ 43ÿ463ÿ9532ÿ5ÿ*2%+8,ÿ6&45663ÿ3ÿ 96564!23456ÿ29ÿ35ÿ%ÿ69ÿ%36ÿ35ÿ546ÿ 529"ÿ86ÿ349ÿ2963ÿ57.ÿ3ÿ5*2%ÿ3451'ÿ 49ÿ34ÿ%ÿ3ÿ*2%+8,ÿ52"ÿ,6ÿ3ÿ5ÿ 59ÿ49ÿ9323ÿ46ÿ*2%+4ÿ546ÿ4632)ÿ3ÿ34ÿ 94627ÿ49ÿ963ÿ35ÿ5*2%ÿ3451'ÿ34ÿ463ÿ&42ÿ2/8ÿ 2"ÿ'ÿ34ÿ94627ÿ93239ÿ3ÿ546ÿ46ÿ23ÿ #532"ÿ $5ÿ3ÿ594327ÿ6&45663ÿ3ÿ2963ÿ57.ÿ49ÿ 4.ÿ56ÿ23)ÿ4ÿ62%79ÿ35ÿ2ÿ43ÿ2947ÿ5ÿ56ÿ 23463ÿ35ÿ2653"ÿ6ÿ$4ÿ3ÿ49ÿ4&6ÿ%473ÿ2963ÿ 57."ÿ 8"ÿ:#6*'#ÿ 'ÿ2963ÿ57.ÿ29ÿ4937ÿ393ÿ43ÿ946ÿ 2&ÿ35ÿ69ÿ323ÿ3ÿ35ÿ ÿ529ÿ2ÿ 546ÿ "ÿ8639ÿ5ÿ5*2%ÿ3451'ÿ26ÿ 4!5734ÿ94627ÿ579ÿÿ5663ÿ35ÿ3ÿ946ÿ2&ÿ 6235ÿ533"ÿ'ÿ623ÿ94627ÿ6ÿ29ÿ0!"ÿ 'ÿ546ÿ59ÿ29ÿ9323ÿ46ÿ*2%+8,ÿ52)ÿ ÿ $4"ÿ-"ÿ*2%+8,ÿ52ÿ53746ÿ5ÿ94627ÿ ÿ26ÿ 546"ÿ$5ÿ294ÿ ÿ&ÿ94627ÿ329ÿ29ÿ2;932%7ÿ 4739)ÿ94627ÿ2743ÿ2239ÿ26ÿ94627ÿ2743ÿ464235"ÿ ÿ $4"ÿ3"ÿ7473ÿ.46327ÿ2963ÿ57."ÿ ÿ $4"ÿ0"ÿ.46327ÿ9739ÿ5ÿ965659ÿ393"ÿ/<ÿ#46ÿ2&)ÿ 4ÿ49ÿ5ÿ43ÿ23ÿ#532"ÿ7<ÿ#46ÿ2&)ÿ4ÿ49ÿ =>?5ÿ43ÿ*2%+8,ÿ52"ÿ

64 ÿ 387ÿ79&ÿ ÿ45ÿ!ÿ7ÿ8ÿ397ÿ1392ÿ65ÿ8ÿ8ÿ 59397ÿ2ÿ486ÿ37ÿ65ÿ98ÿ7ÿ3ÿ ÿ 92ÿ59397ÿ1392ÿ4885&ÿ259ÿ3757ÿ5ÿ35!97ÿ5ÿ 53ÿ 979ÿ8ÿ8697ÿ2ÿ7697ÿ1ÿ73365ÿ1392ÿ8ÿ ÿ12342ÿ77ÿ9259ÿ8ÿ3956ÿ 237ÿ181ÿ157ÿ789ÿ!&ÿ92ÿ$79835ÿ434ÿ 97%$:;$7ÿ ÿ12ÿ 12342ÿ5678ÿ7959ÿ92ÿ483ÿ847ÿ3ÿ259ÿ8915ÿ 9282ÿ93ÿ7356ÿ7ÿ86817ÿ92ÿ7356ÿ598ÿ157ÿ ÿ8ÿ9ÿ483ÿ92ÿ73567ÿ1ÿ8398ÿ 863ÿ3ÿÿÿ3ÿÿ37ÿ3ÿ92ÿ7697ÿ8ÿ9237ÿ 98ÿ979ÿ92ÿ579ÿ486ÿ2ÿ7!'49ÿ157ÿ6&3ÿ 979ÿ2ÿ89ÿ63ÿ37ÿ9ÿ9282ÿ73ÿ15ÿ5356ÿ839 3ÿ(ÿ$39ÿ7697ÿ8ÿ8ÿ8697ÿ,-.ÿ37ÿ8!953ÿ1392ÿ 85938ÿ;59ÿ8ÿ<)( ÿ!&ÿ92ÿ$79835ÿ959ÿ 93852ÿ,-./ÿ5ÿ,-.0ÿ5ÿ ÿ8ÿ92ÿ48ÿ 3543ÿ8'49ÿ)/=))0<7)< ÿ5ÿ!&ÿ92ÿ$85ÿ 12342ÿ37ÿ48ÿ1392ÿ259ÿ8915ÿ2ÿ89ÿ63ÿ5678ÿ ÿ12342ÿ5ÿ8!953ÿ1392ÿ579ÿ486ÿ ÿ838ÿ9282ÿ92ÿ$85ÿ>3856ÿ:689ÿÿ 87ÿ9282ÿ73ÿ15ÿ5356ÿ ÿ5!"#$%ÿ85ÿ#9ÿ37ÿ373!6 5ÿ563ÿ5ÿ92ÿ7&797ÿ5ÿ7&42887ÿ1392ÿ542ÿ892ÿ 2ÿ379ÿ397ÿ1ÿ453ÿ89ÿ8ÿ8ÿ8697ÿ ÿ9259ÿ48ÿ73567ÿ ÿ12342ÿ37ÿ48ÿ ÿ >$$>$7$ÿ 405ÿ#ÿK /5ÿ>ÿ*ÿ392 ÿ?ÿ8 ÿ?ÿÿ*3 ÿ*ÿÿ& 81ÿ5ÿ92ÿ73567ÿ1ÿ379ÿ8ÿ5!89ÿ()ÿ7487ÿ #ÿ53938ÿ92ÿ4ÿ5797ÿ1ÿ453ÿ89ÿ1392ÿ 4U5ÿ?ÿ ÿ8ÿ899356ÿ ÿ B =< 0))Jÿ 9356ÿ*67ÿ%5ÿ"68439&ÿ3ÿ*59397ÿ1392ÿ885&ÿK59ÿ:3757ÿ 8ÿ:39ÿ39&B ÿ/iijÿ ÿ9ÿ357 ÿlmneopgoÿrsÿlots ÿÿ"3355 ÿ86ÿ/= ÿ9ÿ39713 ÿcdefgh ÿ8ÿ0 ÿ86ÿ= ÿÿ/(<+/<( ÿÿ= ÿ98ÿ8!953ÿ92ÿ*%"ÿ3ÿ5895ÿ 55ÿ*%"7ÿ1ÿ ÿÿ3ÿ(ÿ5ÿ3ÿ 7697ÿ8ÿ&ÿ869ÿ5ÿ3ÿ92ÿ*%"7ÿ2ÿ,-.ÿ37ÿ57ÿ!&ÿ73ÿ93852ÿ2ÿ,-./ÿ157ÿ ÿ57ÿ92ÿ5938ÿ8ÿ92ÿ37954ÿ956ÿ!&ÿ92ÿ67ÿ 15ÿ8ÿ6!81ÿ98ÿ3ÿ3ÿ5ÿ92ÿ93ÿ65&ÿ!91ÿ 92ÿ918ÿ67ÿ157ÿ2ÿ,-.0ÿ37ÿ ÿ733656&ÿ98ÿ 2ÿ73567ÿ1ÿ ÿ3ÿÿ5ÿ92ÿ 45ÿ9ÿ*369 4=5ÿ9ÿ*369 2& ÿ579B ÿÿ"3355 ÿ,dvmpewsÿxygmszÿ86ÿ0j ÿ?ÿ9531 ÿ>ÿ95953 ÿÿ>/+>ui ÿ:ÿ9553 4(5ÿÿ9ÿ!!57 [\X]Lÿ,FE^SÿXY_PNYFGOYGOÿ`EOaSÿEOÿXY_Sÿ]PEWSÿLOTSÿ`EbcdNS ÿ86ÿ 93ÿ65&ÿ!91ÿ67ÿ157ÿÿ3ÿ(ÿ37ÿ373!6ÿ9259ÿ *"%7 2,-./ÿ5ÿ,-.03,-./ÿ!&ÿ73ÿ37954ÿ!91ÿ1ÿ5ÿ!3ÿ98ÿ5ÿ92ÿ 85ÿfÿ65&886 0I 356ÿ*8473ÿ873ÿ$ ÿ>4ÿ593ÿÿ 8!ÿ369ÿ8ÿ*67ÿ:65&ÿ579B 8ÿ/ ÿÿ=)(+=)i ÿÿ<j+i= ÿ9ÿ357 ÿ>ÿ575 ÿ0)/)ÿ ÿ0))iÿ ÿ>ÿ97 ÿ,deoe^gf_petfgcdvÿeptogwÿ,fe^ympot ÿ?ÿ9531 ÿlmneopgoÿrsÿlots ÿ86ÿ/( ÿ 57ÿ*%"ÿ2,-.3ÿ#ÿ457ÿ92ÿ,-.ÿ37ÿ ÿ 92ÿ,-./ÿ5ÿ,-.0ÿ5ÿ3457ÿ#9ÿ48787ÿ98ÿ92ÿ ÿ12342ÿ1ÿ79ÿ5678ÿ3ÿ8ÿ387ÿ7542ÿ ÿ12342ÿ5ÿ ÿ8ÿ92ÿ25ÿ5ÿ6ÿ5937ÿ ÿ5ÿ4855!6ÿ1392ÿ4ÿ34ÿ 2ÿ3956ÿ579ÿ486ÿ8ÿ2& ÿ #"ÿ78#7ÿ ÿ5678ÿ 73567ÿ ÿ157ÿ!369ÿ5ÿ979ÿ98ÿ79&ÿ92ÿ 873!363937ÿ98ÿ3587ÿ92ÿ ÿ3ÿ56&ÿ7953ÿ 2ÿ3956ÿ486ÿ5!67ÿ98ÿ379ÿ92ÿ73567ÿ 8ÿ39ÿ ÿ8ÿ!8&ÿ3463ÿ67ÿ157ÿ8ÿ 32&ÿ5ÿ5937ÿ#ÿ53938ÿ92ÿ2& ÿ ÿ5ÿ833593ÿ8ÿ39ÿ347 7&428876&ÿ2ÿ579ÿ486ÿ157ÿ &ÿ ÿ5ÿ48ÿ ghi ÿ

65 APPENDIX 1 Continued PUBLICATIONS Publication II Pilt K, Meigas K, Ferenets R, Kaik J (2010) Photoplethysmographic signal processing using adaptive sum comb filter for pulse delay measurement, Estonian Journal of Engineering, 16: (DOI: /eng ). 65

66

67 ÿ679ÿ6ÿ378787ÿÿÿÿ12 ÿ68ÿ172ÿ!"#$#%&'$"()*#+,-%"./ÿ).+0-&ÿ%,#/').0+ÿ1).0+ÿ-2-%$.3'ÿ )1*ÿ/#*4ÿ5.&$',ÿ5#,ÿ%1&)'ÿ2'&-(ÿ*'-)1,'*'0$ÿ ÿ ÿ ÿ885ÿ697ÿ9894ÿ:987ÿ;754ÿ97ÿ<8ÿ698=ÿ ÿ ÿ Pÿ\]^_`abc^\`]ÿ 9589ÿ4854ÿd5ÿD94ÿM7ÿ4D6Q7ÿ5D95ÿ88Cÿ84ÿ87F4Gÿ?6?658679ÿ56ÿ4G4568Bÿ 94ÿ>6??ÿ59467ÿ?846B58B9ÿ?4ÿ47464ÿ88VÿefÿCDÿ875F9ÿ M5Q7ÿ5Dÿ?9=ÿ6ÿ5Dÿ:RQ9Fÿ67ÿ5Dÿ3gVÿ97ÿ5Dÿ98487ÿ675ÿ6ÿ5Dÿ88Vÿ 3gVÿ48794ÿBD99B58kÿ5DÿB58B9ÿ9B58F85Gÿ6ÿ5DÿD95ÿ3B58B9ÿQ9F4ÿ 12 ÿ

68 ÿ498ÿÿ79ÿ83487ÿ74ÿ8ÿ638ÿ74ÿÿ8ÿÿ34568ÿ ÿÿ76ÿ68ÿ39ÿ8ÿ8838ÿ 4"ÿ8ÿ838 49ÿ"45ÿ8ÿÿ7438ÿ7ÿ858ÿ4ÿ8ÿ79#ÿ88ÿ8ÿ44ÿ 3349#ÿ59$ÿ78ÿ"4ÿ54949ÿ44ÿ68"749ÿ9ÿ79ÿ%8ÿ463ÿ 7ÿ3987ÿ8ÿ5878ÿ%8ÿ9389ÿ#ÿ3ÿ7ÿ4"89ÿ8ÿ4ÿ9"8#ÿ7ÿ ÿ9897$ÿ3987ÿ8ÿ #ÿ8"838ÿ9ÿ7388ÿ9ÿ8ÿ78ÿ9ÿ44ÿ&9$ÿÿ75ÿ"349ÿ4"ÿ ÿ7ÿÿ ÿ463ÿ839!8ÿ"4ÿ5879ÿ3987ÿ9ÿ44ÿ ÿ7438ÿ9ÿ838ÿ4ÿ79ÿ&9$ÿÿ75ÿ"349ÿ4"ÿ8ÿ8"838ÿ9ÿ 5749ÿ548ÿ+9ÿ8ÿ8"8349ÿ548#ÿÿ ÿ7ÿ638ÿ,389ÿ4ÿ8ÿ %88ÿ8ÿ4ÿ59ÿ$7ÿ4ÿ5878ÿÿÿ79*ÿ8ÿ8"8349ÿ9ÿ8ÿ97 8ÿÿ8ÿ88ÿ4ÿ8ÿ44ÿ"4ÿ9ÿ8ÿ98$9ÿ78ÿ'()ÿ 8ÿ$ÿ8ÿ ÿ987ÿ9ÿ8ÿ9897$ÿ 458ÿ9ÿ 5878ÿ 7$934947ÿÿ8ÿ8ÿ8ÿ9ÿ86897ÿ49ÿ8ÿ3987ÿ9ÿ8ÿ678ÿ 7388ÿÿ7ÿ ÿ%8ÿ.ÿ ÿ4"ÿ8ÿ79ÿ 548#ÿ8ÿ ÿ9ÿ8ÿÿ7438ÿ8ÿ638ÿ49ÿ466478ÿ787ÿ4"ÿ8ÿ 8"838ÿ9ÿ7388ÿÿ9897$ÿ"45ÿ8ÿ79ÿ7"38ÿ+9ÿ8ÿ975749ÿ -878ÿÿ797ÿ39ÿ8ÿ 458ÿ%8ÿ ÿ58787ÿ8ÿ9758ÿÿ9897$ÿ 8ÿ$ÿ8ÿ ÿ%8ÿ ÿ58787ÿ8ÿ 6878ÿ9ÿ678ÿ44ÿ 88598ÿ79ÿ"49ÿ7ÿ8ÿ7687ÿ6ÿ4"ÿ8ÿÿ79ÿ88ÿ8ÿ8 9ÿ8ÿ4ÿ44ÿ 48ÿ89ÿ587ÿ758ÿ9ÿÿ4"ÿ8ÿ.ÿ ÿ ÿ79ÿ768ÿ7ÿ74$ÿ 458ÿ4"ÿ8ÿ8598ÿ78ÿ%8ÿ/ÿ ÿ7ÿ 458ÿ%8ÿ568ÿ4"ÿ8ÿ/ÿ ÿ39ÿ8ÿ $9ÿ9ÿ7ÿ79ÿ649ÿ7ÿ"3ÿ4ÿ 8ÿ94ÿ4ÿ *ÿ8ÿ.ÿ9ÿ8ÿ/ÿ 87ÿ74$ÿ9ÿ8"837ÿ 497ÿ 79ÿ"49ÿ9ÿÿ79ÿ68ÿ'5)ÿ+9ÿ7ÿ38ÿ8ÿ8948ÿ7ÿ8$ÿ7ÿ.(3ÿ 7ÿ55ÿ/3349ÿ4ÿ68 75$ÿ4ÿ2ÿ7ÿ7878ÿ4ÿ8ÿ5878ÿ8889ÿ(34ÿ4"ÿ8ÿÿ79ÿ ÿ4ÿ081ÿ8378ÿ4"ÿ644ÿ68"749ÿ78ÿ9ÿ"889ÿ94787ÿ 08#ÿ648ÿ98ÿ98"889381ÿ9ÿ5449ÿ"37ÿ."889ÿ5847ÿ +9ÿ564ÿÿ79ÿ54949ÿ8 47ÿ8783#ÿ8ÿ678ÿ8$ÿ0.1ÿÿ7ÿ5878ÿ 387ÿ'7)#ÿ8ÿÿ79ÿ39ÿ8ÿÿ4ÿ 8ÿ889ÿ 98"88938ÿÿ7ÿ"89ÿ487ÿ94ÿ ÿ4ÿ854 8ÿ8ÿ.ÿ ÿ9ÿ8ÿ"8!893$ÿ94787#ÿ939ÿ648ÿ98ÿ 8ÿ8ÿ998ÿ94787ÿ9$ÿ79ÿÿ967ÿ"8ÿÿ7ÿ6478ÿ4ÿ 758ÿ9ÿÿ8ÿÿ79ÿ%8ÿ/ÿ ÿ4"ÿ8ÿÿ79ÿ39ÿ 49ÿ789747ÿ9ÿ9ÿ ÿ378ÿ$ÿ5449ÿ39ÿ8ÿ73387"$ÿ #ÿ3ÿ8ÿ9ÿ8ÿ9ÿ67ÿ849ÿ8389ÿ8783ÿ7ÿ74ÿ749ÿÿ 8ÿ"8ÿ'1#:)ÿ%8ÿ"7ÿ96ÿ4"ÿ8ÿ6 8ÿ94787ÿ378ÿ$ÿ8ÿÿ78974ÿ 8ÿ8738ÿ$ÿ7ÿ5493ÿ ÿ<9589ÿ5493ÿ7ÿ88ÿ4ÿ8ÿ ÿ79ÿÿ9478ÿ9ÿ8ÿ78349ÿ96ÿ7ÿ998ÿ9478ÿ"45ÿ338849ÿ ÿ ÿ94787ÿ8ÿ854 45ÿ"8ÿ7ÿ9ÿ89 8ÿ584ÿ4ÿ854 8ÿ$ÿ79ÿ8ÿ;-ÿ45ÿ 8ÿ94787#ÿ3ÿ8ÿ79ÿ8ÿ 8ÿ$ÿ79ÿ49ÿ ÿ4"ÿ8ÿ"9589ÿ"8!893$ÿ9$ÿ79ÿÿ345ÿ"8#ÿÿ837ÿ8ÿ 8ÿ6749ÿ"8!893$ÿ/ÿ8ÿ48ÿ ÿ8ÿÿ8ÿ"8!893$ÿ 8ÿ"8ÿ7ÿ8ÿ ÿ 12

69 334ÿ56789ÿ986ÿ885ÿ98ÿ555ÿÿ8655ÿ8ÿÿ5ÿ ÿ334ÿ56789ÿ65ÿ9ÿ8ÿÿ8598ÿ87ÿÿÿ6ÿ8ÿ 8ÿÿ68ÿÿ8989ÿ8ÿÿ9ÿÿ6ÿ 585ÿ85ÿÿÿ99ÿÿÿ6ÿ4ÿ56789ÿ98ÿÿ5ÿ95ÿÿ 65ÿ96ÿ ÿ ÿa8ÿ65ÿ96ÿ6ÿ65ÿ95ÿ9ÿ9ÿ567895ÿ9ÿ65ÿ6ÿ6ÿ65ÿ8ÿ65ÿ 688ÿ8ÿÿ958ÿÿ3"#2ÿ 55ÿ98ÿ4!8ÿÿ6ÿ5678ÿÿ334ÿ567895ÿ98ÿ98955ÿ65ÿ 59ÿ 9897ÿ56789ÿGH ÿ65ÿ56789ÿbc DIÿÿEÿ65ÿ687ÿ98ÿ5ÿÿ59ÿ8ÿ68ÿÿ DEÿ85655ÿÿ669ÿ9F8ÿ595ÿÿÿ $%ÿ'()*+,-ÿ ÿbc 58ÿ98ÿIÿ65ÿ6ÿ68ÿ585ÿ 687ÿ"8687ÿ8ÿÿ9ÿÿ ÿ6ÿ8ÿ585ÿ85655ÿÿ9ÿ565ÿÿ79ÿ59ÿ56f5ÿ BCBC ÿ5ÿjakÿÿ6ÿ65ÿl5ÿ95ÿbmcÿÿ DEÿ65ÿÿ6ÿ68ÿ56789ÿBC PESDEDEO Oÿÿÿ6ÿ55ÿÿÿ595ÿÿ PEÿ65ÿÿ6ÿÿ56789ÿ98ÿOÿ65ÿ98ÿ QR N ÿgthÿ ÿ[ÿ65ÿ98ÿ687ÿg2tth ÿ56789ÿ98ÿ995ÿÿ997uÿoÿ95ÿ685ÿÿÿ6ÿ8ÿ 585ÿV8ÿÿ!98 9ÿÿ6768ÿ98ÿÿ8ÿÿOÿX5ÿWÿ8ÿ59ÿ86ÿ6ÿ9ÿBMCYÿ WÿBT2Cÿ9ÿÿ6ÿ65ÿÿ6ÿÿ8ÿÿOÿ5ÿ WSL Z[ GNTH Q O\ ÿgnhÿ ÿ`ÿ65ÿÿ896xÿ8ÿ^ÿ65ÿÿ59687ÿ689ÿ98ÿaÿ65ÿÿ ÿ6ÿ8ÿ585ÿ98ÿÿl5ÿ95ÿbmcyÿ 8ÿ585ÿJ67ÿT9ÿ55ÿÿ8ÿ585ÿÿÿ5ÿÿ [O az^ Sÿ68ÿÿGNHÿÿ SRÿ]ÿ556687ÿZ^ GHL5 `SR O^O^ NN Z`` W_` ÿgbhÿ ÿ 12 ÿÿ5cÿ65ÿÿ59ÿ8ÿ 8ÿ585ÿL665ÿ9F5ÿ9ÿ65ÿÿÿ8989ÿ8ÿTcYÿ 88ÿ95ÿÿÿ5ÿÿ8ÿÿXÿ8ÿ 58ÿÿÿ8ÿ585ÿÿ65ÿ6ÿ5ÿ8ÿÿÿ"ÿ 6ÿ6ÿT2 OSÿX5ÿ98ÿJ67ÿTÿ8595ÿÿ56ÿ!98 co T S Wÿÿ5ÿ ÿ6ÿ c ÿgdhÿ 5 ÿ

70 4567ÿ97ÿÿÿÿÿÿÿÿ!ÿ2"ÿ#$ÿÿ%&ÿ#ÿÿÿÿ ÿÿ'()ÿ ÿ %%+ÿÿ+ÿ))ÿ../ÿÿ2/+ÿ%ÿÿÿÿÿ+ÿÿ %ÿÿÿ,%-ÿÿÿ../ÿ)ÿ0ÿÿ%ÿÿÿÿ %ÿÿÿ%)ÿ3%ÿÿ%%&ÿ%%ÿÿÿÿ +ÿ-ÿÿÿ,#ÿÿÿ%ÿÿ&ÿ%)ÿ,ÿÿÿÿÿÿ../ÿÿ)ÿ1ÿÿ+ÿ%%ÿÿ2,)ÿ2+ÿÿ %ÿÿ&ÿÿ-ÿÿÿ%%&ÿ../ÿÿ)ÿ0ÿ ÿ*ÿ%ÿ+ *ÿ%ÿÿ,#ÿ%ÿÿ4ÿÿÿ5)ÿ6)ÿ7ÿÿÿ ÿÿÿ,ÿÿÿÿ,%-ÿÿÿ%ÿÿ ÿÿ:ÿÿ+ %%&ÿ%%ÿ : =ÿ%&-+ÿ+ ; ÿÿ62++ =: >ÿÿ9ÿÿ<ÿÿÿ 8<= ; ÿÿÿÿ&ÿ 8ÿ%%ÿ ÿÿ%%)ÿ*ÿ,#ÿÿÿÿÿ6 ÿ%-ÿÿ%)ÿ0ÿÿ&ÿ%%ÿ2+ 8ÿ%ÿÿ 8ÿÿÿ ÿÿÿÿÿÿÿÿ%%ÿ6 ÿ?ÿ Aÿÿÿ%ÿ../ÿÿ%%ÿÿÿÿ -%-ÿÿ2/ÿ)ÿ2%ÿbcaÿ%4ÿdÿÿÿ 1ÿÿÿÿÿÿ+ÿ9ÿÿÿÿÿ%ÿ)ÿ 9: ; <= ) ÿ@ÿ ÿ ÿÿÿ4ÿ%%)ÿ ÿ 12

71 ÿ4567ÿ97ÿÿÿÿÿÿÿ2ÿÿÿÿÿÿ ÿ ÿ!"#ÿ ÿ$%!ÿÿÿ&"ÿÿ%&ÿÿÿ&ÿ'!ÿÿÿÿ 9ÿÿ:! %&ÿÿÿ#ÿ(ÿÿÿÿÿÿÿ&ÿ &#ÿ)%ÿÿÿÿÿÿ*+ ÿÿ&"ÿÿ/ÿ,.ÿÿ,ÿÿÿ%&ÿÿ#ÿÿÿ%ÿ%&ÿÿ 2,-ÿÿÿÿÿ ÿÿÿ! ;ÿÿÿ< , ÿ*C+ÿ ÿ ;ÿÿÿÿÿ0=1>ÿ 0*+1# ÿ*8+ÿ 5,B!:&ÿÿÿÿ/ÿÿÿ!ÿÿ<ÿÿÿ )ÿ!:<ÿ! DAE5 66-ÿ*1+ÿ '! ÿ Dÿÿÿÿ0=1>ÿ 2E56ÿÿF? 5B 2 *+# ÿÿÿ?ÿ ÿ,6ÿÿÿÿg#ÿh&#ÿÿÿ 12 ÿ!ÿÿg#ÿh#ÿ ÿ#ÿ$%"ÿÿi/#ÿ*h+ *+'! ÿÿ/"ÿ!ÿÿÿ&"ÿ0=1ÿ *+2 2ÿ ÿ KAJ6-- L A5,5J 2 5,J L 5JL L ÿ*=+ÿ ÿ

72 ÿÿ)ÿÿÿÿÿÿÿÿÿÿÿ (.,ÿ0ÿ#ÿÿÿÿÿÿÿÿÿ 5678ÿ98ÿÿÿÿÿÿÿÿÿÿ (ÿ)ÿÿÿÿÿ*+,ÿ ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ ÿ-.ÿÿÿÿÿ/ÿ/%ÿÿÿÿÿÿÿÿ ÿ +,ÿ0ÿÿÿÿÿÿÿ.ÿÿ12,ÿ3/ÿÿ ÿ ÿ"ÿ#$ÿÿ%ÿ"ÿ&'ÿÿÿÿ 0ÿ#ÿÿÿÿÿÿÿ(ÿÿÿÿÿÿÿ+ ÿ(/ÿÿÿÿÿÿÿÿ.ÿ#2,#ÿ-,ÿ ÿÿ4ÿÿÿ%ÿÿÿÿÿ0ÿ#,ÿ5ÿÿÿÿ ÿÿÿ(ÿÿ%ÿ 6789:ÿ;8ÿÿÿÿÿÿÿÿÿÿÿÿ+ÿÿÿ ÿ ÿ #$ÿ#2,#ÿ ÿ=ÿÿ$,$ÿ ÿ2ÿÿ3,>ÿ ÿ>ÿ#=,#ÿ ÿ?ÿ#=,?ÿ ÿ&ÿ##,&ÿ %ÿ</ÿ-ÿ ÿ ÿ 12

73 *&ÿ'ÿ #!#ÿ'ÿ&ÿ'$ÿ*&ÿÿ&ÿ"%&ÿ ("#'ÿ&*ÿ&ÿ&'+ÿ, #ÿÿ(5!'&(ÿ' )'(ÿ#!#'ÿ6007+ÿ4ÿ'!$ÿ# ÿÿ ÿ&&ÿ&ÿ"%&ÿ!''ÿ%'ÿ"!ÿ 3434ÿ6789ÿÿ8ÿ "ÿÿ# $ÿ"%&ÿ#!''ÿÿ(#'ÿÿ'%ÿ')ÿ $ÿ-.'+ÿ/01ÿ(ÿ/21ÿ&ÿ"!&)!&ÿ#%#!%& $& ÿ" $ÿÿ%ÿ!$ÿ +ÿ4ÿ"%&ÿ*&'ÿ 67<;=<>;? 67CDAA $ÿ"%&8ÿ('#(ÿ3ÿ-.+ÿ/218ÿ#ÿÿ EE B "ÿ :; 6/017 ÿ ; BCA; :; = ÿ/0f1ÿ ÿ*ÿ; -.+ÿ/0f18ÿ&ÿ# =ÿ'ÿ&ÿ*&ÿ ') (ÿ".!#3ÿ') LJM "ÿ&ÿ#!#+ÿg3ÿ!'ÿ('#&ÿ, /10COO ÿ-#ÿ*&8ÿ8 '8ÿ'$%%3ÿ& B9A0 ÿ-.+ÿ/h18ÿ'ÿ!ÿ&'" &(Iÿ $ÿ ÿ =ÿ('#'ÿ&ÿ$ BC 0: =K N /01 ÿ ;:; B!&ÿ ; AB9"ÿ" AA ;J? A0$& =N8ÿ*#ÿ'ÿ&Pÿ" ÿ/001ÿ $ÿ 1ÿ&ÿ* &ÿ) ')+ÿrÿ&ÿ# ÿ.!%ÿ& 1ÿ%ÿ*&'ÿ' 4ÿ#'ÿ!'ÿ#!#ÿ" ÿ +ÿ"ÿ*&'ÿ#!'ÿ&ÿ#'ÿÿ&ÿ"%&ÿ".!#3ÿ') ; ÿ&ÿ"%&ÿ!&)!&ÿ#%#!%& +ÿqÿ-.+ÿ/218ÿ%ÿ&ÿ*&'ÿ 'ÿ&ÿ"'&ÿ'(ÿ% &$ÿ&ÿ(5!'&(ÿ"%&ÿ".!#3ÿ') %ÿ'& )U(ÿ$&!(ÿ $ÿ"%&ÿ*&'ÿÿ(5!'&(ÿ##!%(ÿÿ'ÿ'$%ÿ'ÿ) (ÿ& TTTÿ&ÿ&ÿ'$ÿ ÿ&ÿ" % 0S*ÿ#&Iÿ 'ÿ =ÿ/f8 'ÿ'ÿ) ÿ$&!(ÿ VVÿ*ÿ;ÿ(ÿ:ÿÿ&'1ÿÿ#%#!%&(ÿ ;: '%8ÿ'ÿ('#(ÿ3ÿ-.+ÿ/H1Sÿ "ÿ&ÿ "ÿ&ÿ(5!'&(ÿ"%&ÿ' '%ÿ(ÿ09 U(5!'&(ÿ"%&+ÿ 'ÿ' ===A!%(ÿÿ'ÿ#%!%(ÿÿ&ÿ%'&ÿ'ÿ% :'ÿ& ÿ&ÿ"%&ÿ *ÿ %$&(+ÿqÿ&ÿx&ÿ'&)8ÿ&ÿ$&xÿ'ÿ'#(ÿ& '.!ÿ$&#'ÿyÿ(ÿz8ÿ')#&%3+ÿ4ÿ($' (5!'&(ÿ"%&ÿ(ÿ&ÿ'&!$#%3+ÿ4ÿ".!#3ÿ') -.+ÿ/001ÿ" :Aÿ"ÿ9+ "ÿ&ÿ(5!'&(ÿ"%&ÿÿ&ÿ)%#ÿ,%&ÿ*&'ÿ;,'&8ÿ%ÿ&ÿ%!'ÿÿ&ÿzÿ$&x8ÿx#(ÿ&ÿ% 08:TÿQÿ&ÿ#'ÿ[8 'ÿ" =ÿ" ÿ&ÿ(5!'&(ÿ"%&ÿ'ÿ#%#!%&(ÿ" $ÿfÿ& ÿ0ÿ*&ÿ'&)ÿ+wÿ4ÿ$&!(ÿ $ÿ ÿ%ÿ*&'ÿ3ÿ#ÿ; :BÿZÿ(ÿYÿÿ00 )U(ÿ$X$!$ÿÿ#%#!%&(ÿ(ÿ')&(ÿ& "ÿ&ÿ"'&ÿ'(ÿ% WW \ÿ$&#'+ÿ ÿ$x$!$ÿ "ÿzÿ(ÿyÿ.!%ÿ& *(ÿ$x$!$8ÿÿ "ÿ&ÿ U ÿ 12 (ÿ ÿ" >W= _&XÿZÿ%$&'ÿÿ+ Bÿ]#ÿ'&&'ÿ*ÿ "ÿ&ÿ$&xÿ [+ ÿ&ÿ#'ÿ28 :Bÿ*#ÿ$'ÿ&&ÿZÿ'ÿÿ[^ÿ$&X+ÿ *+ÿqÿ&ÿx&ÿ'&)ÿf8 aÿq&ÿ$!'&ÿÿ ÿ &(ÿ&&ÿ!+ÿ]#ÿ'ÿx)%(ÿ Ẁ= 08JW= `J> BBBÿ4ÿ%!ÿ F+ `Bÿ0 JBÿ(ÿÿ&ÿ>ÿ%!ÿ'ÿ "ÿ>ÿ'ÿ#(ÿ!&%ÿ&ÿ 9

74 88638ÿÿ3889ÿ6ÿ48ÿÿ867ÿ5ÿ58ÿÿ69ÿ434ÿ5ÿ6ÿ ÿ6ÿ48ÿ869ÿÿ48ÿ5ÿÿ48ÿ8ÿ85689ÿ356667ÿ ÿÿ*ÿÿ+ÿ4ÿ5ÿ8ÿÿ**, 36689ÿ6ÿ48ÿ874ÿ58ÿ69ÿ 86589ÿ88ÿ569ÿÿ49ÿ3869ÿÿ48ÿ3696ÿ 8ÿ48ÿÿÿ8838ÿ8863ÿ868ÿÿ59"89ÿ8ÿÿ786ÿ6ÿ 538ÿÿÿÿ569ÿ ÿ48ÿ59"89ÿ8ÿ58ÿ69ÿ569ÿ356667ÿÿ6489ÿÿ48ÿ#ÿ5ÿ58ÿ6ÿ 07ÿ,ÿ %835ÿ35356ÿÿ48ÿ874ÿ88ÿ598ÿ6ÿ&'(')ÿ!874ÿ88ÿ ÿÿ8ÿ456ÿ48ÿ589ÿ5ÿ$488ÿ356667ÿÿ -.ÿ58ÿ/ÿ4ÿ874ÿÿ59"89ÿ ÿ!874ÿ ÿ ÿ :ÿ 23456ÿ78ÿ953589ÿ874ÿÿ48ÿ59"89ÿ3ÿ8ÿ +<ÿ <ÿ / =<ÿ <ÿ, ÿ 2ÿ*>1ÿ*>*ÿ*=>ÿ*+/ÿÿ 48ÿ358ÿIÿJÿ1KÿF5Gÿ8863ÿ868ÿ6ÿ358ÿ:ÿJÿ=LÿFGÿ:ÿJÿ,LÿF3Gÿ:ÿJÿ2ÿ BCD8ÿE8ÿ'9"89ÿF9ÿ68Gÿ569ÿ66H59"89ÿF95489ÿ68Gÿÿ3ÿ8ÿ8863ÿ868ÿ6ÿ ÿ ÿ 12

75 *+9ÿ((4ÿ!5" ÿÿ 3454ÿ7898ÿ9ÿÿÿ9ÿ8ÿÿÿÿ8ÿ!"#ÿ ÿ. $%&'ÿ()*+ÿ!ÿ,!!"5ÿ:.%"/ÿ&;ÿÿ/1ÿ((4ÿ!5" -.#ÿ$/0"ÿ/1ÿ234ÿ!5" ÿ1 ÿ%0ÿ$1 ÿ67 &!%-.'ÿ!/ÿ!ÿ 8ÿ "#ÿ/1ÿ %:ÿ/1ÿ((4ÿ!5" #!:!<-/ÿ/%ÿ#/.,!"ÿ/1ÿ/,!"!,-,ÿ%!"/'ÿ$-/ÿ!/ÿ!ÿ"%/ÿ-:!<!"/>ÿ./!"5ÿ%!"/ÿ%:ÿ/1ÿ((4ÿ!5"./!"5ÿ%!"/ÿ%:ÿ/1ÿ.<-."<;ÿ="ÿ.&!%-ÿ/-#!ÿ/1ÿ ÿ.<-."<ÿ0 <<-. ÿ#/</#ÿ:.%,ÿ!/ÿ,!!"5ÿ:.%"/ÿ!ÿ/1ÿ:!,-,ÿ%.ÿ //ÿ.!%#ÿ-#ÿ!"ÿ/1ÿ:!/.ÿ< <1.<-."<ÿ :%.ÿ &.,'ÿ!/ÿ!ÿ%!$ÿ/%ÿ!"/.% 3%,$ÿ:!/.ÿ%-/-/ÿ< "5!"5ÿ ÿ,%.ÿ.<!ÿ#/.,!" ÿ />;ÿc!/1!"ÿ%"ÿ.<-."<'ÿ/1ÿ,.ÿ#/</#ÿ %ÿ$ÿ/ /!%"ÿ%:ÿ()*+;ÿb1ÿ*+9ÿ&ÿ!ÿ< "#ÿ/1ÿ*+9ÿ&ÿ!ÿ< ;ÿ="ÿ< /!%";ÿ /ÿ/1ÿ!5" 8"ÿ ÿ/1ÿ< ÿ!ÿ# %ÿ <- ÿ/1ÿ:.%"/ÿ.!ÿ!" ÿ.%#-</ÿ%:ÿ/1ÿ,%&!"5ÿ0!"#%0ÿ >#ÿ:%.ÿ1 /#ÿ*+9ÿ&ÿ!ÿ$/0"ÿ/0%ÿ <-!,-,ÿ /#;ÿ :ÿ%:ÿ/1ÿ"-,$.ÿ%:ÿ "#ÿ,!"!,-,ÿ%!"/ÿ <-.>;ÿ=/ÿ /#ÿ:%.ÿ %0ÿ <1ÿ 01.ÿHÿ!ÿ/1ÿ"-,$.ÿ%:ÿ.<-."<ÿ!"ÿ/1ÿ((4ÿ!5" B1ÿ()*+ÿ,!ÿ</#ÿ/1 E/ÿ/1ÿ<%,$ÿ:!/.ÿ%-/-/ÿ$ÿ< /ÿ()*+ÿ!ÿ "ÿ&. 5ÿ%:ÿ/1ÿ/0%ÿ<%"<-/!&ÿ.<-."<ÿ()*+;ÿ <- /#ÿ:.%,ÿ/0%ÿ<%"<-/!&ÿ.<-."<;ÿ=/ÿ "-,$.ÿ%:ÿ.<-."<ÿ-#ÿ!"ÿ/1ÿ:!/.ÿ%-/-/ÿ< "ÿ$ÿ0.!/"ÿ "ÿ& -ÿ!ÿ< ()*+FG H/#ÿ K HH IJ ()*+FG HP KQ ' <- ;ÿ()*+@ÿ!"#ÿ@ÿ, /!%";ÿÿ5".!l /!%"ÿ%:ÿ "ÿ/1ÿ ÿfd@gÿ 2M- 0!"#%0ÿ"5/1ÿ;OÿB1ÿ%-/-/ÿ%:ÿ/1ÿ,%&!"5ÿ!"5ÿ/%ÿ!"-/ÿ$>ÿ1 D ()*+FG; OOKIJ PHO H ÿfdrgÿ SBETÿ/%ÿ < &. <- ="ÿ /!%"ÿfdrgÿ< 5ÿ%&.ÿOÿ.<-."<ÿ /!%";ÿ=/ÿ, ##!/!%"ÿ/%ÿ/1ÿ.%<#-.ÿ#<.!$#'ÿ//ÿ!5" ""ÿ$ÿ!"/../#ÿ >ÿ/1ÿ!":-"<ÿ%:ÿ/1ÿ<%,$ÿ:!/.ÿ%"ÿ/1ÿ()*+ÿ, "ÿ/1 :ÿ/1ÿ0!"#%0ÿ"5/1'ÿ01!<1ÿ!ÿ!,! /ÿ/1ÿ()*+'ÿ, "#ÿ# ÿ/1ÿ,%&!"5ÿ >#ÿ$>ÿ1 -.#ÿ:.%,ÿ/1ÿ<%,$ÿ:!/.#ÿ!5" ÿ:ÿ%:ÿ/1ÿoÿ.<-."<;ÿ D&. 5ÿ0!"#%0ÿ!ÿ# 5ÿ0!"#%0ÿ,/1%#ÿ0!/1ÿ/1ÿ ÿ0.ÿ5"..ÿ/%ÿ/1ÿ:!/.ÿ%-/-/ÿ >#ÿ.5 -.,"/;ÿ 'ÿ!ÿ /#ÿ!"ÿ.#7 12 /1.%-51ÿ.< / B1ÿ:!./ÿ!"-/ÿ%:ÿ/1ÿ!5"!5" ()*+ÿ/1 =,-ÿ ÿfu!5;ÿ*<g;ÿ2 /ÿ#/.,!"ÿ/1ÿ((4ÿ!5".#ÿ <1ÿ-"!/ÿ!,-ÿ, /ÿ<%"/ ÿ5". "/ÿ:.m-"<>ÿdÿvl;ÿb1ÿ<%"#ÿ5". /%.ÿ!ÿ/1ÿ-"!/ÿ!,-ÿ/..8ÿ/1ÿ234ÿ!5" ÿ.!!"5ÿ:.%"/ÿ# ÿ67 >ÿ:.%,ÿ/1ÿ67!"'ÿ. 8;ÿ/#ÿ/%ÿ/1ÿ234ÿ FU!5;ÿ* ()*+ÿ& 8"ÿ <1ÿ5". ÿ G;ÿ.!ÿ$/0"ÿ+;@*ÿ ÿ/, /#ÿ.<-."<ÿfu!5;ÿ*$g;ÿ="ÿ/1ÿ,!##ÿ%:ÿ/1ÿ5".!"5ÿ!/ÿ:%.ÿ&.>ÿ5". /;ÿb1ÿ((4ÿ!5" "#ÿ+;r*ÿ;ÿw"ÿ.!%#ÿ:.%,ÿ /#ÿ.<-."<ÿ/%ÿ ÿ/, /%.ÿ!"-/ÿ!ÿ 8ÿ:%.ÿ ÿ/ÿ0 ÿ/./<1#ÿ <1!&ÿ/1ÿ5!&"ÿ()*+ÿ 0ÿ((4ÿ!5" "#ÿ<%,.#ÿ /#ÿ!5" ÿ0 'ÿ/1ÿ

76 ÿÿ()#ÿÿ*+",ÿ 5678ÿ98ÿÿÿÿÿÿÿÿÿ ÿÿ$%&ÿ'ÿ!ÿÿ#ÿÿÿÿÿÿÿ ÿÿ-ÿÿ.ÿÿÿ/$%ÿ #ÿÿÿ ÿ ÿ,ÿ*ÿÿ ÿ!ÿ,ÿ1 ÿÿ0ÿÿÿ ÿÿ2,ÿ3,ÿ-ÿ!ÿÿÿÿÿ/$%ÿ ÿ ÿÿÿ.ÿ!!ÿ'ÿ!ÿ!ÿÿ"ÿÿ!ÿÿ#ÿÿ ÿÿÿÿ!ÿÿÿ ÿ 5678ÿ48ÿ5ÿ/$%ÿ'ÿ 0ÿ'ÿÿ'ÿ,ÿ!ÿÿ!ÿ ÿ.ÿ!!ÿ'ÿ!ÿÿÿÿ/$%ÿ'ÿ'ÿ ÿÿ ÿÿ#ÿÿÿÿÿÿ ÿ ÿ0ÿÿÿ ÿÿ#ÿÿÿÿÿÿÿÿ ÿ 12

77 ÿÿ!ÿ"ÿÿ##"ÿÿ'ÿ""ÿÿ"$%ÿ&ÿÿ 4567ÿ87ÿ9ÿÿÿÿÿÿÿÿÿÿ!ÿ"ÿÿ##"ÿÿÿ""ÿÿ"$%ÿ&ÿÿÿ ÿÿÿÿ ÿ (ÿ+ÿ"ÿ%!ÿÿ"$%ÿ&ÿÿÿÿ&ÿ&$ÿÿ ÿÿ(ÿ,ÿ!ÿÿÿ"$%ÿ&ÿ&"ÿ&ÿ"ÿÿÿ!ÿÿÿÿÿ(ÿ-ÿÿ$ÿ&ÿÿÿÿ!ÿÿ ÿ"ÿ#.ÿ&$ÿ()ÿÿ(*ÿ(ÿ/ÿ0ÿÿÿÿ&ÿÿ ÿÿÿÿ"ÿÿÿ$ÿ&ÿÿÿ&$ÿ()ÿÿ(*ÿ(ÿ &ÿÿ'ÿ""ÿÿÿÿÿ&ÿÿ#ÿ"$%ÿ&ÿ ÿÿÿ'ÿ""ÿ&ÿÿ&ÿ#ÿ""(ÿ ÿ$$ÿÿÿ"ÿ&ÿÿÿÿÿÿÿ"$%ÿ&ÿÿ -ÿÿ"ÿ&ÿÿ#ÿÿ"!ÿÿ"ÿ$ÿ&ÿÿ#ÿ"$%ÿ ÿ ÿ""ÿ!ÿÿÿÿ"ÿ%(ÿ/ÿ&ÿ"ÿÿ ÿ!ÿ$ÿ%ÿÿ'#"(ÿ ÿ 17ÿ :ÿ;8<ÿ52:=>9:ÿ ADLÿÿÿ%M"ÿÿÿNÿÿÿÿ"ÿ#"(ÿ %.ÿ%ÿ""(ÿaÿÿ"ÿ.".ÿÿÿb%cÿ?'#$ÿ""ÿÿÿ.ÿ&ÿÿ#ÿ"$%ÿ&ÿÿÿ ÿ ÿ()ÿÿ)ÿhiÿÿÿ&ÿ1ÿÿ*!ÿ#".(ÿ+&ÿ%g#ÿ ÿ#"ÿÿ"ÿÿÿ9+,b+k(ÿ,ÿÿÿ&ÿ"ÿÿÿ "ÿÿ$ÿÿ*ÿhiÿ$#ÿÿÿjfg%ÿ(ÿÿ #ÿgÿÿÿg#ÿ&(ÿ/-lÿg#ÿÿg#ÿ&!ÿ ÿ%.ÿÿÿ"ÿÿ$!ÿÿÿ"g&ÿ&n"ÿ /.!ÿÿÿÿÿ&ÿÿÿ%g#ÿ&ÿ"$#ÿ&ÿÿ 11 ÿÿ%"ÿ&ÿnÿÿ/(ÿ1(ÿ+&ÿgÿÿg#ÿ&!ÿ "$%ÿ&!ÿ"ÿÿ&ÿ""ÿ&ÿÿ#ÿ""(ÿ Aÿ.ÿÿÿ$ÿ$#%ÿÿ"ÿÿ#ÿÿ(ÿ,ÿ,ÿÿÿÿÿ/(ÿ1(ÿ,ÿÿÿÿÿÿ"%.ÿÿ ÿ

78 ÿ564ÿÿ84ÿ""#ÿ59ÿ95ÿ959!4ÿÿ48478!4ÿ ÿ95ÿ5ÿ9ÿ9ÿ14ÿ ÿ59ÿ5448ÿ564783ÿ95ÿ5ÿ9ÿ9ÿ1ÿ943ÿ984ÿ645ÿ 6643ÿ8ÿ ÿÿÿÿ3ÿÿÿ 4ÿ59ÿ843ÿ84ÿ6894ÿ7!ÿ9843ÿ95ÿ94ÿ9ÿ9ÿ17ÿ 4ÿ943ÿ984ÿ64ÿ3ÿÿÿ66435ÿ 4ÿ ÿ59ÿ5448ÿ 4ÿ65894ÿ %&'(ÿ*(ÿ+,ÿ-ÿ""#ÿ59ÿ34734ÿ3ÿ84ÿ344.ÿ+!,ÿ""#ÿ59ÿ843ÿ9/ÿÿ/65ÿ ÿ+7,ÿ""#ÿ59ÿ843ÿ6894ÿ7!ÿ9843.ÿ+,ÿ59ÿ564783ÿ843ÿ9/ÿÿ/65ÿ9843.ÿ +4,ÿ59ÿ564783ÿ843ÿ6894ÿ7!ÿ9843ÿ ÿ ÿ 12

79 ÿ4944ÿ4ÿÿÿ378ÿ4ÿ34567ÿÿ4ÿ ÿ97ÿÿ7345ÿÿÿ48ÿ7645ÿÿÿ8ÿÿ43ÿ 63ÿ4ÿ$ÿ46438ÿ4ÿ6#47*ÿ5ÿ34634ÿ5ÿ43ÿ34ÿ!"ÿÿ9434ÿ ÿ94ÿ4ÿ6#47ÿ9ÿ73ÿ6ÿ4ÿ$ÿ ÿ9ÿ6474ÿÿ4ÿ7ÿ43ÿÿ4ÿ 2ÿ463448ÿ4ÿÿ5ÿ 4ÿÿÿ9ÿ4345ÿ9ÿ.ÿ5ÿ9.ÿ43ÿÿ547345ÿÿ4ÿ ÿ6ÿ4ÿ7ÿ4345ÿÿÿ 745ÿ ÿÿ9434ÿ9ÿÿ,-+8ÿÿ4ÿ6#47ÿ9ÿÿ563ÿ 4643ÿ%&'(ÿ4ÿÿÿÿÿ4)434ÿ9ÿ344345ÿ3ÿ3445ÿ 346ÿ4)434ÿ4ÿ/ÿ4ÿ4)ÿ48ÿ4ÿ!"ÿÿ+ÿ40ÿ5ÿ4ÿÿÿ3ÿ3ÿ,6464ÿ4ÿÿÿ 4ÿ4)434ÿ"3ÿ9ÿ54ÿ4944ÿ9ÿ 3ÿ 9ÿ4345ÿ9ÿ9ÿ7ÿ438ÿ97ÿ645ÿ9ÿ5ÿ63ÿ ÿ3ÿ4ÿ 43ÿ4ÿ34735ÿ48ÿ4ÿÿ5ÿ!"ÿÿ9434ÿ37445ÿ4ÿ 2ÿ46344ÿ9ÿ54ÿÿ6ÿ39ÿÿÿ5ÿ4ÿ475ÿ4ÿ 2ÿ9ÿ46345ÿ4944ÿ4ÿ 2ÿ46344ÿ4ÿ 66ÿ776ÿ1ÿÿ4ÿ4345ÿÿÿ9434ÿ645ÿÿ4634ÿ 2ÿ 8ÿ4ÿ434ÿ 34634ÿ/ÿÿÿ46ÿÿ743ÿ5ÿ34634ÿ4ÿ 4643ÿ7ÿ4ÿ44ÿÿ&2 736ÿ!"ÿÿ/434ÿ 434ÿ64ÿÿ 4ÿ346ÿ34ÿ9ÿÿ2ÿ1ÿ7ÿ43ÿ34ÿ34ÿÿ4ÿ$ÿ 2634ÿ17ÿ9ÿÿÿ4ÿ4ÿÿ4ÿ$ÿ4643ÿ41 28ÿÿ6ÿ5434ÿ ÿÿÿ5ÿ4ÿ ÿÿÿ2ÿ15ÿ2634ÿ137ÿ9ÿ4ÿ77645ÿ 2ÿ65ÿ4ÿ04ÿÿ344ÿ5ÿ ÿ58ÿ4ÿ 4ÿ58ÿÿ6ÿÿ34ÿ643ÿÿ ÿ3ÿ43ÿ4ÿ8ÿ34ÿ 7ÿ5ÿ34634ÿ748ÿ97ÿÿÿÿ743ÿ44ÿÿ43ÿÿ 2ÿÿÿ4ÿ4944ÿ4ÿÿÿ2ÿ1ÿ5ÿ67ÿ4ÿ 43ÿ58ÿ4764ÿÿ43ÿ4ÿÿ74ÿÿ4ÿÿÿ4ÿÿ/ÿÿ 4ÿÿ63445ÿ4ÿ5447ÿÿ4ÿÿÿ3ÿÿ34ÿ3474ÿ8ÿ4ÿ 4ÿ7ÿ43ÿ 43ÿ4ÿÿ4ÿ43ÿ9ÿ8ÿ7645ÿÿ4ÿ43ÿ343ÿÿ 27ÿ3ÿ3ÿ5447ÿ3ÿ5347ÿ3445ÿÿ5ÿ34634ÿ74ÿ8ÿ4ÿ 4ÿÿ676ÿÿ4ÿÿÿ2ÿ1ÿÿ4ÿ7645ÿÿ3474ÿ 2ÿÿÿ544345ÿ3ÿ4ÿ63ÿ ÿ43ÿ7ÿ43ÿÿÿ 4743ÿÿ5ÿÿ74ÿ3ÿ4ÿ643ÿÿ ÿ 5ÿ5545ÿÿ4ÿÿÿ263ÿÿÿ9ÿ5434ÿ,-+ÿ9434ÿ44345ÿ 9434ÿ47ÿ64ÿ3045ÿ4ÿ4ÿÿ4ÿÿÿ435ÿ 64ÿÿ5434ÿ643ÿÿ ÿ9ÿ4ÿÿÿ9ÿ44345ÿÿ <6ÿÿ4ÿ4ÿ9ÿ44345ÿÿ6ÿ4ÿ:;1ÿ35ÿ643ÿ4433ÿ /ÿ558ÿÿ ÿÿ9ÿ443458ÿ745ÿÿ6ÿ648ÿ :;1ÿ4ÿÿÿ9ÿ44345ÿÿ6ÿÿ4.435ÿ4ÿ44ÿ 44345ÿÿÿ43ÿ34ÿ34647ÿ9ÿ345ÿ3ÿ&ÿÿ<ÿ=>ÿ563ÿ /ÿ63ÿ4)434ÿ94ÿ7345ÿ5#645ÿ5ÿ.5#645ÿ7ÿ43ÿÿ 12 7ÿ43ÿ5#645ÿ43ÿ9434ÿ6ÿ94ÿ4ÿÿ4ÿ<ÿ43ÿ438ÿ 44345ÿÿ9434ÿ4345ÿ9ÿ344ÿ5#645ÿ5ÿ344ÿ.5#645ÿ54ÿ 4ÿÿ,-+ÿ9ÿ46345ÿ5ÿ4ÿ46ÿ9ÿ77645ÿ ÿ

80 -ÿÿÿÿ +ÿÿ ÿ4567ÿ97ÿÿÿÿÿÿÿÿÿÿÿÿ ÿÿ!"#$ÿ%ÿÿ&'(ÿ ÿÿ ÿ%ÿÿ)*ÿ ÿ!!(ÿ,ÿ-ÿÿÿ ÿ!!(ÿ+ÿÿ ÿ-ÿÿÿ ÿ ÿÿÿÿÿ ÿ*ÿÿ%ÿÿÿÿÿ*ÿ.ÿ/ÿ ÿ/ÿ0ÿ-ÿÿÿ ÿÿÿÿ!"#$ÿ%ÿÿ&'(ÿ ÿÿÿÿ+ÿ 2ÿ-ÿÿ ÿ1.ÿ2$ÿÿÿÿÿ.ÿ/ÿ ÿÿÿÿÿÿÿÿ ÿÿ.ÿ ÿÿÿ ÿÿÿ ÿ!!(ÿ,ÿ ÿÿÿÿ ÿÿÿÿÿÿÿÿ*ÿÿÿÿÿ /ÿ3+ÿ2$$4ÿ* ÿÿÿÿ*ÿÿ-ÿ-ÿ2.ÿ5ÿÿÿÿÿÿ ÿÿÿÿ-ÿ*ÿÿÿÿÿ ÿ89:;<ÿ=7ÿ>) 7ÿÿ*ÿ.ÿ ÿ ÿ-.ÿ6ÿÿÿ,ÿÿÿÿ ÿ ÿ 2 6ÿÿ,ÿ4ÿ2BAÿ33Bÿ3CDÿ ÿ ÿ 2?ÿÿÿÿÿÿ 62 0ÿAÿ#ÿ ÿÿÿÿÿ*ÿ >ÿ,ÿ 6ÿÿ,ÿ4ÿ0$$ÿA$$ÿ#$$ÿ >) Eÿÿ3Aÿÿ03ÿÿ01ÿ Eÿÿ2Bÿÿ3Aÿÿ03ÿ 2 ÿÿÿ ÿ 12

81 )ÿ ÿÿÿÿ&ÿÿÿÿÿ ÿÿÿ 5678ÿ98ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ ÿ!"#$ÿÿÿÿÿ%ÿÿÿÿ ÿ ÿ!'#$ÿÿÿÿÿ(ÿ ÿ ÿ!#$ÿ ÿÿÿÿÿÿÿÿ5ÿ6ÿÿÿÿÿ -.ÿ/ÿÿ-0ÿ123ÿ ÿÿ.ÿ0ÿÿÿÿÿÿÿÿÿ *ÿ +ÿÿÿÿÿÿÿÿÿÿ)ÿ,ÿ ÿ)ÿ4ÿÿÿÿÿ*ÿ%ÿÿÿ ÿ ÿÿ )ÿ4ÿÿÿÿÿÿÿÿÿ7ÿÿÿ ÿ8ÿÿÿÿÿ ÿ8ÿ9ÿ:)ÿ -ÿÿ-ÿ)ÿc00ÿÿ-00ÿ-ÿÿÿ ÿÿÿÿÿ ÿÿcdeÿÿ-ÿÿÿÿÿÿÿÿÿÿ ÿ ÿÿÿÿÿÿÿÿ-)ÿ1/ÿÿ 7I0ÿÿ0ÿÿ-ÿ)ÿ*ÿÿ-0ÿÿ0ÿ ÿÿ 0ÿÿÿÿÿ-ÿÿÿ-00ÿÿÿ0-ÿ0ÿ -0)ÿ *ÿ-ÿÿÿÿÿ ÿiÿÿÿ0ÿÿÿÿffeÿ0ÿ.ÿ ÿ ÿÿ0ÿÿÿÿfg(h)ÿ*ÿ ÿÿÿÿÿ ÿÿÿ 12 00ÿÿ.ÿFFEÿ0)ÿ*ÿÿÿ 00ÿÿ)ÿÿÿÿ/ÿÿ7.ÿÿ ÿÿÿÿ ÿ)ÿ*ÿÿÿ ÿÿ-0ÿ-ÿÿ)ÿ-ÿÿÿÿÿ ÿ)ÿ ÿÿ.ÿÿiÿÿÿÿÿÿÿÿ ÿÿÿÿÿÿÿ ÿ+ÿÿÿÿÿÿÿÿÿ ÿÿ'ÿÿ ÿÿÿ-ÿ ÿÿ.ÿ ÿ 0ÿ-ÿ ÿÿ

82 ÿ53686ÿ45ÿ ÿ9ÿ45ÿ ÿ845ÿ97474ÿ95ÿ965ÿ95ÿ ÿ436ÿ3ÿ ÿ845ÿÿ786ÿ45ÿ35ÿ675ÿ9ÿ57565ÿ 45ÿ9ÿ845ÿ35386ÿ5368ÿ5386ÿ36ÿ3ÿ!%&'ÿ84ÿ55ÿ3635ÿ 57565ÿ86ÿ96434ÿ49ÿ45ÿ ÿ845ÿ45ÿ6985ÿ ÿ55ÿ 573ÿ4ÿ45ÿ35ÿ485ÿ45ÿ3745ÿ845ÿ75ÿ394ÿ485ÿ5ÿ ÿ 84ÿ454ÿ863ÿ3ÿ5ÿ3ÿ84ÿ595ÿ863ÿ9ÿ45ÿ754ÿÿ45ÿ)333ÿ 9ÿ45ÿ33485ÿ9ÿ845ÿ96ÿ45ÿ!%&'ÿ3ÿ3635ÿ(5ÿ9ÿ845ÿ ÿ 36575ÿ(5ÿ35386ÿ554ÿ9ÿ45ÿ9ÿ845ÿ97ÿ5ÿ43*56ÿ8649ÿ39764ÿ 9ÿ5897ÿ57565ÿ436ÿ45ÿ ÿ845ÿ8ÿ 6985ÿ434ÿ35ÿ45ÿ35ÿ36ÿ84ÿ84ÿ45ÿ!!"ÿ863ÿ36ÿ45ÿ3886ÿ964ÿ 35ÿ545435ÿ(5ÿ!!"ÿ863ÿ3ÿ5ÿ3ÿ45ÿ!%&'ÿ ÿ8ÿ3545ÿÿ ÿ!%&'ÿ ÿ 8455ÿ!!"ÿ36ÿ#$"ÿ863ÿ55ÿ75ÿ49ÿ5375ÿ45ÿ!%&'ÿ(5ÿ867565ÿ #79536ÿB6896ÿ497ÿ45ÿ#79536ÿC58963ÿ%559564ÿ ;9ÿ<&'=>ÿÿ45ÿ#496836ÿ43545ÿ863686ÿ954ÿ9 ÿ(8ÿ47ÿ3ÿ7945ÿÿ45ÿ#496836ÿ98565ÿ ÿ:364ÿ ÿÿ D2E2D2.,27ÿ 76ÿ ÿ?ÿf3ÿgÿh583ÿiÿi338ÿ%ÿi3438ÿcÿi38*ÿgÿ36ÿc9366ÿhÿ$ ÿ99ÿ575ÿ ÿ2ÿ3ÿcÿ56549ÿÿ(997836ÿgÿf37564ÿ!ÿ!36685ÿÿ83ÿÿhÿ(354ÿcÿ36ÿ ÿaÿf3ÿgÿh583ÿiÿi3438ÿcÿi338ÿ%ÿi38*ÿgÿ36ÿc9366ÿhÿf483ÿ36ÿ55483ÿ ÿ786ÿ5585ÿ786ÿ75ÿ35ÿ43684ÿ485ÿ ÿJ6ÿKLMNOÿQRSTÿ F5ÿÿJÿ5564ÿ9ÿ34583ÿ ÿÿ ÿ75ÿ35ÿ5984ÿ UVVWXYÿZVS[LVXS\MVXYÿ]MV^[L[VN[ÿM^ÿST[ÿZ ÿ_v`oÿa[boÿc\myoÿdn\oÿ936ÿ 549ÿ9ÿ75ÿ35ÿ43684ÿ485ÿ ÿ36ÿ84ÿ ÿ84ÿ34583ÿ99ÿ ÿ=ÿ!84ÿiÿh583ÿiÿf3ÿgÿc9366ÿhÿ36ÿi38*ÿgÿ63975ÿ4545ÿ%$ÿ996564ÿ ÿ&ÿi33ÿÿq365ÿgÿj86ÿ"ÿ36ÿh536ÿÿ9*86ÿ ÿeÿ3ÿ585ÿ ÿ_v`oÿa[boÿc\myoÿdmnoÿ I388ÿÿH369ÿqÿ;33ÿHÿ54ÿ3ÿ!75ÿ43684ÿ485ÿÿC35345ÿ8635ÿ949 ]MtlWSOÿa[STMbgÿKLM`LOÿc\Mt[bOÿ?1o1ÿmuÿA&<eA=1ÿ 5493rÿC585ÿ9ÿ45ÿ ÿ36ÿ5963ÿ558565ÿsOÿ]Y\VOÿaMV\SOÿ ÿoÿ$9498ÿ"ÿh56596ÿwÿ36ÿc37*3ÿ!ÿÿ933485ÿ ÿ9ÿ33485ÿ6985ÿ ÿ<ÿv99ÿfÿÿ36ÿ33ÿqÿf9ÿ38365ÿ33485ÿ845ÿ9ÿ36586ÿ94896ÿ34834ÿ86ÿ ]MV^[L[VN[ÿA''<ÿ=&Ae=&&ÿ 5335ÿ ÿ569ÿ863ÿJ6ÿKLMNOÿZ ÿ_v`oÿa[boÿc\myoÿdmnoÿ ÿ3984ÿ9ÿ8688p86ÿ94896ÿ34834ÿ86ÿ3ÿ9553ÿ97645ÿ5335ÿ 75ÿ98545ÿJ6ÿKLMNOÿZ ÿ_v`oÿa[boÿc\myoÿdmnoÿ]mv^[l[vn[ÿa''<ÿ?&aoe?&2?ÿ ]MV^[L[VN[ÿF896ÿA''<ÿ?''=e?''1ÿ ÿ1ÿ$766863ÿ#ÿ!ÿx\`\sxyÿy\ys[l\v`zÿuvÿzvslmbwns\mvÿgÿv85ÿ;5ÿw9*ÿ?11&ÿ ÿ 12

83 335ÿ142ÿ55ÿ6878ÿ55ÿ ÿ5ÿ9ÿ5ÿ5ÿ:82892ÿ;ÿ72<8ÿ8ÿ;=ÿ428ÿ;ÿ11>ÿ 345ÿ677898ÿ5ÿ5ÿ8ÿ5ÿ5ÿ9ÿÿ5ÿ5ÿ 3F5ÿ142ÿ55ÿ6878ÿ55ÿ5ÿK5ÿ9ÿ5ÿ5ÿ11>ÿ8794ÿ7;8897ÿ;ÿ748ÿ84Lÿ;7297ÿ=Lÿ ÿ?9ÿ ÿ34ÿ31115ÿ 897ÿ72<8ÿ;=ÿ4285ÿAMBC0ÿ+ 3I5ÿP828ÿQ5ÿ5ÿ7298ÿR5ÿ8ÿ5ÿS5ÿ9ÿS;9<829;ÿ5ÿ5ÿQ828ÿ;ÿL7;<;48ÿ I4YJI3I5ÿ ÿz[\[]^_\`ab[cdeefg^ga_ÿagciee^gÿ\jj\^kaÿ]k^agÿlg\_emeÿ ÿ582:9ÿ142ÿ54:ÿ6878ÿ9ÿ ÿ:ÿqÿ5ÿ bnn\bga_^ÿeoe]\gla_ÿpebbfg^\dgceÿ 748ÿr11>sÿ8794ÿ2u8<ÿ;9ÿ<84ÿ98228ÿ748ÿ<28:85ÿQ8498t ÿ<28:485ÿ ÿ:482ÿ<82ÿ8482;;748ÿrq5>sÿ8794ÿt77ÿ:ÿ;2;7482q8;t 87944u8ÿ88828ÿ28845ÿQ5>t8794ÿt77ÿ2v82<ÿ7ÿ11>t <288ÿ8288ÿ;9ÿ2v842ÿ848ÿ;8488<82ÿ<88u75ÿP<4ÿ98t 8794ÿ;8ÿ47825ÿ?7ÿ42882<ÿ;8ÿ:;8ÿ9287;48828ÿ88498<ÿ 92ÿ2ww8ÿ2ww22ÿ<v4:ÿ72<98ÿ428ÿq8ÿ4288ÿ8ÿ82=ÿ Q5>t8794ÿ ÿ:ÿ94qq82ÿ848ÿ42ÿu:ÿ11>t879448ÿ997ÿ ;8ÿ8487ÿ<u885ÿv4:9ÿ<284ÿ<u828ÿ782ÿ78;;82ÿ<8t 84ÿ ÿ11>t8794ÿ<u=ÿ;4ÿu:22ÿ4882ÿ2972ÿq8825ÿ xww8ÿ84882ÿ72<8ÿ42ÿ<v4:9ÿ<228ÿ;<28ÿ 882ÿ8ÿ;ÿ<v8ÿ9;2ÿ<v4:9ÿ<288ÿÿ;92ÿ4285ÿ 82<vv28ÿww9ÿ;82882ÿ<v8ÿ475ÿ 848ÿ<v4288ÿ;92ÿ422ÿ9ÿ82ÿ888ÿ<284ÿ828ÿ 8794yqt88ÿÿ<v88ÿ28ÿ87948ÿ<u<ÿ882u2ÿvÿ5ÿ 2<ÿ:v:8;9=7ÿ82<vv28ÿ:ÿ988ÿ ÿ 8988ÿ88822ÿ;828ÿ<ÿ;9ÿ<u4ÿ8<2ÿ88ÿ 9422ÿ28822ÿ ÿ11>t87947ÿ4848ÿ;4ÿ482ÿ:84ÿq5ÿ 8ÿ82<ÿ428ÿ8788=ÿ874842ÿÿ<v8ÿ;8ÿ82<ÿ4285ÿ 5;92ÿ428ÿ88ÿq8ÿÿzÿÿ<v8ÿÿ;92ÿ428ÿ 8ÿ;828ÿ<ÿ:885ÿ x79888ÿ282ÿ ÿ828ÿ;92ÿ428ÿ88498<2882ÿ78;;t 498ÿ<28:ÿ<2845ÿx82ÿ282882ÿ8478ÿ82ÿ748498ÿ<2887ÿ;9ÿ9t 22ÿ7;;48ÿ;828ÿ<ÿ<u5ÿ148498ÿ<28:ÿ28ÿ74ÿ;8ÿ;t x8828ÿ2488ÿ9928ÿ82ÿ424ÿ;9ÿ8882<ÿ882ÿÿ748t 12 2ÿ11>t8794ÿ2u8<ÿ;925ÿ 4882ÿ2972ÿqÿ8ÿ8<ÿ11>t8794ÿ878878;;98ÿ:ÿvvt 528ÿ84828ÿ879487ÿ9v28ÿ82ÿ427ÿ;9ÿ<u4ÿ884ÿ ÿ

84

85 APPENDIX 1 Continued PUBLICATIONS Publication III Pilt K, Ferenets R, Meigas K, Lindberg L-G, Temitski K, Viigimaa M (2013) New photoplethysmographic signal analysis algorithm for arterial stiffness estimation, The Scientific World Journal, vol. 2013, Article ID , 9 pages (DOI: /2013/169035). 85

86

87 *+,+-./01.23/ &'3()31)')%%''#$% 19!"#$%$4 KBD?;LCG8D=;MNOCDGP6B6G6;?MNKC=LQR6DAC?MNSCB?TUVBCGSDGWX6BAMY ]^+_-.2`+a2bcd3b`+e3/-4fag3a+.3aghi+/0ab`+e3/j`hi-43aaka3l+.,32mbci+/0ab4bgmhf032-n-2+2+oh]pqrsi-43aahf,2ba3- t^+_-.2`+a2bcd3b`+e3/-4fag3a+.3aghu3avwb_3agka3l+.3,2mhor]rou3avwb_3aghxy+e+a z1{42619 }1{42)19~8)7) ˆz1{ ) ƒ31& ) )ƒ ) 444} Š Š )! " ) }19 13) 123( Ž Š Ž"66 54 Š Š Š Š )Œ Š š ) Š Š 72 Š# Š0 5Š72351Š š œ {ž Š )Ÿ Š Š ) Š)#) { ) !54Š7238)%$)31 254Š N G;B:WQE;D:G Š Š3142 1Š )69 41 Ž66 54 Š ) Š Š71212}12) Š Ž ƒ" 511ª23 12Š Ž6" % )6"423 ƒ"4 2113Š2)91113Š ƒ"} š412š " Š ) « Š # ) Š31 222&"423 Š Š4 % ) ) Š42ŠŠ ŽŸ1774 ) ŠŠ Š Š $ ) Š66! Š {ž4) 1249Ž" }4 454 Œ412123( Ž ' /'-e'-423+' )49173Š Š95& '-/'-e'-423+'-) Š ) )" Š Œ4 Œ8 Œ Œ3 423Œ Š4 9ŽŸ

88 * ÿ 02!"ÿ##$ÿ%&' +.- +,- *1* +/-13 ()##$ÿ%&' 4%5ÿ *1* * +-7-, 13 4%5ÿ * 9ÿ 02!"ÿ##$ÿ%&' :- *1* ()##$ÿ%&' 134%5ÿ02 3 *1* * 134%5ÿ02 3 ;<=>?@AB1286C2DDEF5C6657FF24625G75G2H57I5F75647HG2FJKLJMKLJ4KLJKL6J2KN0OPQ2RS6T0PQ2PRMSFMU247FV bc572fn7i2hg2fv12wxy5ff272f45m27i2 45GF4C2N7I2FMU247V M27H22645GF45FeN47F67I23fDDE 6b5g2bc572FG2FIFM2266Qg2F775F754Q haijaklvmb5g2bc572fn3fdde6wxy46m2 Ccb272FNF422656Cb27I F2F WXYZR[\]\^\_SàLHI227I2aL[L]L^L6_27I2 567I2F75n62FN7I27252FhATlV M2262C5F722NbioPQ2PI27IQFMU247LH57IWXY d624267cm54756fl7i fi5c G2FL5FFIH656;5C20V123fDDEF5C65Fc42F2L 67I2HG2bc572F C7F7Q MQp5F227VhAolV12F5b5c42F56Cb27IIF M226F2F45M2562F275567I2F752FhA0jAklVd7 5FFb27I77I245GF4FQF72b2F674I6C2 12Fbc2F2Cb267NDDE63fDDEF5C6LHI54IIF G2FI7c25F564F2FNI27IQFMU247FVd75FG5F5M2 Nb;5C207I77I2WXYG2FN7I2I27IQFMU MQI5CI2N7I28F767I5c25FV12 5n22642M27H226bq5b6b565bWXYG2F5FrVOoL HI54I46F75772FM7istNb7I2HI2F42NWXY 2F7FNF56C2FMU247V N7I2c2eFNb ha0lv;7i2b2l7i c2ef567i28f767i5 5n2267cIF2FNDDEF5C6564F2N46F2475G2c25F c25f247277i2m2c56656cnfqf754cif2n7i2 5F277I256Fv45267Fcc2F56NDDEF5C6I5CI2 DDEF5C64bc277I2F2466N7Ic25FVuF 4bc6267F665F2V 2F77I2bc572FN272472c2eF567I246F2475G2 c25f2f45m25n2267cif2n7i2ddehg2nb6 WXYG2FM24b267542MQ5n2267V15F2F7I5CI2 F762G5756NWXY67N7Q5672c27756N7I2

89 ÿ!"!# $$$$!$%$%!$&$&!$'$'! ()*ÿ # $$$$!$%$%!$&$&!$'$'! ()*ÿ # +,ÿ!'!% -.!"!# /01* /01* /01* /01* :;<=>?@A01BCD1B1EC1569F6G1HHIB4E5JDD1D6KF9CLMN1M9G16GNBOP136546BB139514Q64Q1J9R1D6KR46G RQ1D1SB5TUVQ1BW012XHHIB4E54BD931B1Y56G1RQ1CD461B E69B6NONZ4B1 16W[\L]W 01TUVGB69O13361R46G9RB651Q ^ G1BOP136BR46G4531B1B64^51B F9C6G1G16GNBOP136BW_56G4BB6NYR1GQ14CD9Q16G1 2XHHI5NB4BC16G O6456G1TUVQ1B R46GC454CB651Q G1RQ1B6 6G1BC BR46G45951D1499F6G1HHIB4E5W01 E946GC4B61B6195E9D9FG16GNBOP136B5BC E9D9F4O161BD64156BBD496B6NW àbcdefgh ijkjlmnnutopqrststquvwtxyzj{9c F6G1D149}B 15E6GYQ1E45EY76145EY F6G1RQ1BF9C 6G12XHHIB4E514B66145O93S4EC45 :4E1W01HHIB4E54B7611R46G9RM5G4EGMDB :_~761B459169B1D61X39CD95156B5G4EG F1 153N594B1W01369^F BF96G19RM5G4EGM DB761B1B11361B 5 Wƒ Y1BD1364Q1NW 96G761B11B4E51B45E6G1R459RC16G9YR46G6G1 CC45ER459RF536495RG116G1391BD9545E761 91B13G9B15Bƒ F96G19RMDB 15 F9 6G1G4EGMDB761W 2OB1 156N6G1HHIB4E54B4^ R95 F964C1BJ:4E1KW01B4CD1B64^ B 6G14^11531BO16R1156R9395B1364Q1BCD1B9F6G1B4E5Y RG43G4BB9S59R5B6G17B6M4^115314^ W04B S459F4^ R9SBBG4EGMDB761Y56G1G4EG F B1CD471B1B6W 9R1Q1Y6G15R B14B93616G4EG1F BF96G1HHIB4E5WX1 696G11B D1Q49BNY6G12C996G{94B1~9OB6 X4^ J2{~XK4BB1W 012{~XGBO1151Q19D1F94^11563B1B6G6 1D643NO151734F93N45E961 D14C156BR46G 594BN6RG114^ B1 41[\ˆ]W04B4^1M B3G1C1D9B1B1B6G1F99R45E3G3614B643BAD1M 34B169RF BYBC996G5E5611BDD1B495 9FG4EGF BW01919F6G14^ C451B 6G1BDD1B4959F6G1G4EGF BW_56G4BE946GCY6G1 7 G919F6G14^ BB1YRG43G4BB96G19R1B6 D9B4O1951W 66G19R1F BJ Š\ƒ KYRG116G1 CP946N9F6G1D9R19F6G1HHIB4E54B9361Y6G17B6M 4^115314^ {~XF1 153N1BD95B1B1 D3643N1 W _5D36431YO49B4E5BB3GBHHIYRG43G G16364Q46NY11345EO6596D14943W04BC15B6G6 6G1GC954339CD95156B9F6G16R9395B1364Q B 9F6G1HHIB4E5546B14Q64Q1B35O164^1156 F BW_56G4BB6NY6G19RMDB7614BB1R46GB6643

90 )=ÿ%%+ÿ/!$ "#!$!ÿ%ÿ&'( >"!ÿ,$!?*!-2;ÿb6ÿa< )=ÿ+ÿ/!$!ÿ "#!$!ÿ%ÿ&'( ÿÿ &ÿ$! +ÿ,$ &ÿ$! +ÿ,$ &ÿ$!!!-!"ÿ$!-*$!-! )*$ +ÿ,$!ÿ 31!$ÿ,ÿ$!-*$!-!!!-!"ÿ$!-*$!-!.ÿ+ÿ/!$ 31!$ÿ,ÿ$!-*$!-!!!-!"ÿ$!-*$!-!.ÿ+ÿ/!$ 31!$ÿ,ÿ$!-*$!-!.ÿ+ÿ/!$ 3!1!ÿ,ÿ$ÿ,$ÿÿ $!,!$!-!ÿ+ ÿ42ÿ*ÿ567!ÿ,ÿ$!-*$!-! 3 9:;ÿ94:;ÿ9-:;ÿ9":;ÿ"ÿ9!: (!!-ÿ,ÿ1!8ÿ!ÿ,ÿ$!-*$!-! 3 <!$ÿ-$ÿ+ (!!-ÿ,ÿ!ÿ,ÿ$!-*$!-!,$ÿ!1!$2ÿ$!-*$!-! 0+ÿÿ-$!"ÿ*,$ÿ!1!$2ÿ$!-*$!-! 0+ÿÿ-$!"ÿ*,$ÿ!1!$2ÿ$!-*$!-! 0+ÿÿ-$!"ÿ* CDEFGHIJK4L5MNO7P2Q5M6R42Q56MO7P2Q24625S75S26TQ5QU 2M2O2V264TUW4456MTX427566NY2QOPN654 4NR6267Q2RQ267P2Q2QRR2Q2U12 26M7PQO7P2ZZ[Q5M Q27P266N5\2 726Q27P77P2PN6544NR6267Q2R42Q256 7P2QN2]T^C5M2I_U 7P2Q M7PQ5Q`QX]P54P42QR6Q7 7P2RQ2O2V264TO`a\Ub67P5Q4Q2X7P2O6N267 O2V264T5QQ57727`a\UW7P27P24NR6267QT7 7P2O2V264TN75R2QO7P2O6N267O2V264TUb67P2 62c7Q72RX7P2Q5M65Q8722]57P7P2`a\]5276Q5756d Y6Ze872f`ghU12Nc5NN]Y22QO7P2 RQ6Q7RY6Q2Q277iUiìU122QNR56M6 12ZZ[Q5M65Q2QNR256Q4P]T7P762O 87256M2Q4527]57P7P2Q2466O7P 25S75S2QO7P2ZZ[Q5M6UWj287256MX7P24RTO Q Q5M62]57P7P26N5\ QON7P5QZZ[Q5M6U QO ZZ[Q5M6Q46Y25M62YTQ56M5k22675Q75647R567Q ON7P2Q5M6Q2O22642XO2cNR2X7P Nc5NN565NR567O7P25Q56MO67U N565NR56746Y25l477272N562XY24Q2O7P2 ]ST2656MO7P25Q72RPQ2Ub75QQ5l N5627P2ZZ[Q5M6Nc5NR567Q572R26Q67P2 Q772O7P245SQ4QTQ72NfmihU12niR242672S2 O7P2ZZ[Q5M65Q56MO675QQ2Q7P22O22642R567 O7P25M6N267O7P QUC7P2N2X7P2Q246 6O7P25S75S2Q2NS24456M77P2NS2N267 O7P2ZZ[Q5M QU R2S5QT24527Q2R72TO2S2T ZZ[Q5M6QU12S2M2]S2ONO62QYo247]57P57Q g242642q5qm5s2656c5m20u 2O6ON7P2S2M23rZZ[]S2ONUC5Q7TX7P2 122QNR56MX87256MX65M656MR42Q2Q7562 3YQ2V267TX7P2R2LQO]S2QpqXpYqXp4qXpqX6p2q \24Q56MQO7P2S2M2O7P25S75S2]S2ON 2O6U12R2LQO]S2QpqXpYqXp4qXpqX6p2q2 Y27]226\24Q56MQO7P2O7P25S75S2]S2ONQ 575Q2S2256C5M20U3246TX7P2N565N6Nc5N R567QO7P23rZZ[]S2ON2472Y27]2267P2\2 4Q56MQO7P2O7P25S75S2]S2ONU12246Y2 ]S2ONQO7P23rZZ[X]P227P2R2LQO7P2p4q6 pq]s2q67rr2ub67p5q4q2x7p2p4q6pq]s2q 2272N562567P2R42QO7P23rZZ[]S2ONX]P22 7P2O7P25S75S25QNc5NN565NY27]226\2 4Q56MQU Q6S2M2]S2ONO7P23rZZ[2 k2472yt7p22m2o2v264to7p2ze]drq872u 12R75N2M2O2V264TO7P2Ze]dRQ872]Q R75N5\256274P52S27P2]2Q7Q762S5756 O7P23rZZ[]S2NR572QX]P54P75N72TN565N5\2Q 7P2Q762S5756O Ub65756X7P2S ststuvwxyxz{wx } ~ ƒ{ x w ˆ Š ˆ Œ }Ž t 7P2R42N267O]S2QpqXpYqXp4qXpqX6p2q675N2N56 PQ7Y2N565N7PMP77P2R25QO62QYo247U a22x575qqn27p77p245sq4qtq72n2q67 4P6M2S2QP7R25Q564Q2QOP27PTQYo247QU12 R75N5\756O7P2Ze]dRQ8722M2O2V264T]Q 45276Q5M6QONMROP27PTQYo247QU `a\67p22m2o2v264t]q4p6m2y27]22606 `0a\]57PQ72RO`a\U122O7P242QR656M Ze872]QImnn7QNR56MO2V264T`La\U12I PN6542RQ27PMP7P2872Q7P QO 7P23rZZ[]226N5\277P2O2V264TO`a\Ub67P5Q ]TX7P O24PPN6544NR62677]S2QpqX ÌPN6544NR6267Q P2O6N267 12]57PO7P2Ze]dRQ87276Q5756dY6]Q pyqxp4qxpqx6p2q46y26t\2u

91 (.8.+ (.8.+ ( ÿ ##$ÿ %.,./.011.%1.,1./ %"ÿ"!" &ÿ'ÿ##$ÿ 23ÿ(+ ÿ()*##$+ 46:67%1.%.1.%.,./.011.%1.,1./,&-ÿ" &ÿ'ÿ##$ÿ 23ÿ(+ 47 4; <76 45.%.,./.011.%1.,1./ ÿ 23ÿ(+ =>?@ABCD01E1F1GGH546II139514E64E15J96K 14E64E1LE1J9MINO3PO9451QL46K M4R IN6K45451IQS I14F F69 TU9J6K1GGHI4F54I45FJ95656K14I64536LE1IVWXVOWX V3WXVWX5V1W OYI45F6K1R1939I45FI9J6K1J96K 14E64E1S J1d153YS=96K1I651E X6K159e M4R1MZ461IXfXJ9M59M4R12gGGH131531I [\]X^\]X_\]X`\]X5abcL113361J913K1F1 5J9M6K1E1F12gGGHLE1J9ML11I1S01 59M4R1MZ461IJ9M6K1E1F12gGGHLE1J9M L116P15IE1F1E1IfEFS01I651E46495I L113361J9I4F5IJ9M13KK16KYIOh136XiXOY 01I651E46495IJ96K159M4R1MZ461IX LK11w4I6K15MO19JZ1495{4I6K16965MO1 9JZ149I456K1Z931I1I4F5S24M4YX6K1I65 I45FJ99L45F1d6495D 1E464959J6K LE1Z1PI956K164M19M45LI 3361S01I651E46495IL11E1F19E16K1 F9Z9JIOh136IOYI45FJ99L45F1d6495D jklmnopqrstuvflwnxfefyz {x } N~Q LK114I6K16965MO19JK16KYIOh136IS jkefo ƒtujklmn} N Q ˆ Š Œ ŒŽ_ ˆ]Œ ` Œ 014MZ9E12gGGHI4F5 5YI4IF946KMLI61I61956K1I4F5IJ9MF9Z9J K16KY54O161IZ64156IS019Z64MG 9LeZI761 1F1J1d153YLII1J96K15YI4IS012gGGHLE1I L abcE1IL113361L46KI65 1E46495IS 6K1Z96939OY6K K43I 9MM I13K66K1š6495 5I64661J9œ16Kg1E19ZM156X I6954S01GGHI4F5IJ96K15YI4IL111F4I611J9M K16KYIOh136I54O161IZ64156IS 69žžY1IS01YL Z1M5156M K1YL11145FL46KE49I1E1I9JZKYI43364E46Y45 Ž[ `^Œ 01I6YLIZ1J9M1 1ZZ9E9J 6K141E1YY4E1IS I6K1LE1J9M9J6K1GGHI4F5E41I L46KF1X6K1IOh136IL114E K1J99L45FF1 F9ZID TŸ F9ZX1 31Z66K1F9ZI9JTŸžT5žTŸ TX39MZ4I17E1 IOh136IS09I1F1F9ZI39MZ4I16K11IOh136IXO13I1 6K1IOh136I456K1K16KYF9ZL11F1J9M ~ 46LI K16KYIOh136I69J S 5 X ~K16KYIOh136INo QZ6434Z61456K1I6YS 4F59I4IJ9MM S 5X T4O161IZ64156I Z6434Z61456K4IZ496I6YS01Z64156IL11F1 IOh136I456K1F9Z9J4O161IZ64156IK1314E1 TX TŸCTXCTŸTXTŸžTX5žTŸ TS 3KF1 O16L115 5žžY1IS014O161IZ64156IMYKE1 4531I1614I64 51I1696K1I319643Z931I1I456K1 E1I1IS c Œ Ž ` Œ]Œ I4F5Z931I45FLI ªS01K4FKe59LeZI IL OYI45F6K1V7~WJ536495S I1Z61J LIL4615J933645F I9J6K12š g«~ S I9JG 761L113361I45FJ536495I V7ZM9W5V7ZMSW OYI45F51 Z14M156M1I1M15639MZ1 «~ XLK43K 4531š13975F134ZI15I95OeO46GGH M91XM95F96K11E431IS01GGHI4F5LI4F464R1 L46KG e~že~63d4i f4i611l46k O 815E495M156S01IMZ45FJ1d153YLI~PœRS 01~eM456195FI4F5LI1391X5~eZ14995F I1FM156N{o QLII11361J96K12gGGH5YI4IS01 01GGHI4F5IL111F4I611J9M6K F1 ±²³ µ µ I4F51F4I66495LI34196XLK416K1IOh136LI45 01F1513K5F145I651E464959J6K159M4R1 1I645FZ9I46495S01IOh136LI451I645FZ9I I6 MZ461I5abc453I1I9J F1J1d15341I ~TM4561IZ49696K1M1I1M156IS0199M61MZ161 4I4I66145=4F1INQŸN1QS=913K1F1J1d153YX LI95 1F11I45F6K11 Z14M156IS 6K1F4E15I651E464954IE1F19E16K1F9Z9J K16KYIOh136IS01M454ME1F1I651E46495 J9abcX[\]X^\]X_\]X5`\]4ITSTžXTST XTSTCXTST TST 1IZ1364E1YS=9ZM161IX6K1M454MI65 X5

92 ÿ!"#$%&'ÿ(!ÿ)*ÿ+,-"ÿ./ (!ÿ678 9: ÿ!"#$%&'ÿ(!ÿ)*ÿ+,-"ÿ./ (!ÿ;<= 9>: ÿ!"#$%&'ÿ(!ÿ)*ÿ+,-"ÿ./ (!ÿ?<= 94: ÿ!"#$%&'ÿ(!ÿ)*ÿ+,-"ÿ./01 9: ÿ!"#$%&'ÿ(!ÿ)*ÿ+,-"ÿ./ (!ÿA<= 92: BCDEFGHI12J2K2L762J5756L9MNJK:OPQR66S5T2SU572L75V22672K2O2W26452LOXYZ27Z[L>\247L] 9:PQRL75V22672K2O2W26452L^9>:6S5T2SU572L_à75V22672K2O2W26452L^94:6S5T2SU572Lbà75V2267 2K2O2W26452L^9:6S5T2SU572Lcà75V22672K2O2W26452L^92:6S5T2SU572Ldà75V22672K2O2W26452L] 2J5756eLO6eZ227Z22K2O2W264[O7Z2fg872 5L0hT676L5756>65LYhT] 35S5[i56B5K2L09:j092:i7Z2L762J5756L 74Z4725T27Z25LU2L56OeJ2U2kLlmil>mil4milmi 6l2m567Z275S2S562K5J26]12K5J26L76 2J5756L2J2K2J27Z2KUOZ27Z[L>\247L] 12S565SJ2K2L762J5756LOeJ2U2kLlmi l>mil4milmi6l2m567z275s2s562x]xsliy]nsli o]0slix]psli6h]qsli2lu2475j2[]12s565sl76 2J5756L46>2O6O7Z22K2O2W264[O0hTO ej2li2r42u7oej2l>m]s67z24l2oej2l>mi7z2s565s L762J5756eL72K2O2W264[OohT] B7Z2UUL2O4SU5L6i7Z2LS2fftL5K6L e22lu42l2e57z7z2k57zs2l45>2>[g5u L227]vYwx]12J2K2L762J5756OPQR J25Lq]YX] 12PQR6K22756LZ5UO7Z2Z27Z[L>\247L6 5>272LU75267Le57ZL762J5756>L25L772 56B5K2w9:]12fg8722K2O2W264[676L5756u >6e220hT6YhTi2LU2475J2[ieZ54Z4456K7 7Z2U2J5L[U2L26722L7LL22SL7>2U75SO 7Z23yfft6[L5L]s65756i2K2L566[L5LeL L75S727Z22756LZ5U>27e226 PQR6K2>[L56KK2625T2562S2]zL8L7 UU4Z7Z2K262562S2eLL2ieZ54Z5L4L2 O7Z2K2625T2562S2e57Z526757[56kO64756] zoe56k2k2l56s2eluul2i{zts q]qyn}~ Y]HH0]y2LU572O275J2[L5SU2S2Z5KZ42756 ƒ elo6>27e226pqr6k2o7z2z27z[ KUieZ54ZLZeL7Z2L76K LZ5U>27e226 7eJ5>2L] s6b5k2w9>:5746>2l2267z2 6uz7S6U7O 7Z2UUL2S2{zts]12L762J5756LO7Z2 S2MNzts q]yx0]b5>272lu75267l7z2s26pqr J25V22642OS7Z2UUL2S2{zts5L ˆ y5 q]šhn]12pqr5v22642los7z2uul2s2{zts e224su2>27e226z27z[65>272lu75267lkul] f52 u72l797eulsu2ls56k62wj5642l:el U2OS256g3Œr42e57Z q]qh]12l5k j2 OU52 u72l7elž iez54zlzel7z25v22642 >27e2267eKUL]

93 +,-./ ÿ0-.ÿ1-2ÿ'34* +,-./ ÿ0-.ÿ1%2ÿ'34* VWXÿ'-U#U* ÿ 56!"#$%&ÿ'()* ÿ 586!"#$%&ÿ'()* ÿ 536!"#$%&ÿ'()* ÿ 56!"#$%&ÿ'()* ÿ 516!"#$%&ÿ'()* U U U _àc _dae _daf _àb Yÿ!/#Zÿ[ Yÿ!/#Zÿ[ Yÿ!/#Zÿ[ Yÿ'&-!4* Yÿ!/#Zÿ[ jyklhà`dim_dannf Yÿ!/#Zÿ[ ghàid U U U U U 56\]-7^4ÿZ-^]$^4 _àb _àe opqr\]-hàsni Yÿ'&-!4* +,Yklhtàdef +,Yklh_àdef 586 +,-./ ÿ0-.ÿ12ÿ'34* +,-./ ÿ0-.ÿ172ÿ'34* +,-./ ÿ0-.ÿ12ÿ'34* 9:;<=>?@01A1B1C651A46495C5DEAB69F6G1HA1CIJKI8JKI3JKIJK5I1J956G164L19L4564M11561B1F1N15341C F1N15341CT536C651A464959FHA1I3J64M11561B1F1N15341CT56C651A464959FHA1IJ64M11561B1F1N15341CT 516C651A464959FHA1I1J64M11561B1F1N15341CS F9OPG16GQC8R136CS56C651A464959FHA1IJ64M11561B1F1N15341CT586C651A464959FHA1I8J64M11561B1 VWX_jYkl'-U#U* 9:;<=>@5601uvw6x9456CH46GC651A464955yz68CF9B9xC9FG16GQC8R136C548161Cx64156CS01451L91 451{ }~4C395C6361F9B9x9FG16GQC8R136CS5865 6L5x96F9395C6361L91S846G G1C651A A1CF9B9x9FG16GQC8R136C1B4A15S846GCG14514CB4A156G1L15uvw4M11531F9L451L91F948161Cx64156CS

94 !"5"#57$722#2 " %72&'(25")*)+$,5446"75-772".7/0%72,2"42%&'(1234*527 72#57%65"227*1274$722#2" "7,542,2*8"2"7$".9247",575642"2 725"75:62"46.222"5!5: % 27!".9247"$ %45"45"2" * !5# ,"%6.27,226&'(6#2.!"56#72#57, ,226&'(6#26#27!".9247"," ",6*1222"75"225756"%722#2" $B8 C$, "2.!72547% 45" "672".9247D".5#54#2* C E "$"4".2"2*F,22$75" ".246" "42%%72"752"* !5#2&'(2"$ !#%".9247"$,22%6%5.272"75267" <=5#27<>*12"2.255""5"5.256=5#27<.>* =722$72"775"754!"5# :22642,"%6.27, !".9247"65.272"75267"*12 5#2&'(2"24"2.!725642"2725 "75:62"%5.272"75267"*G22722"$"2% "75267"2"55&'(2"42727! ".9247"*C746.24"2.!722!5#6"5"%5.272" 257"$,545"%,2,572H !$6" 2"7272"75:2656#%727252"".226"72* C746.2"226%=5#2"/<>I/<2>7772,2"7 2#2" ,"4522,26722#2 %2;264!%68725"+FA*5"27+FA$7265A ""7772"2.2*12#2" " 65#22#2%2;26452"24"2.!7265"26 6,672%2;264!46267"%72 "5#6$,54 2"577275#2%2;26452"65827# 5: *15"4"2"72%7! %72,2" %72"56#265A #23,2%*87,22#2%2;26452"$ "2"2"2$,54%,2"JK$J.K$J4K$JK$ 56"56#265A "*8"2"7$72572 %72,2"56"56#265A "62#2 3,2%25:2267$,544" "2 56" <=5#2"/<>I/<2>*15"26"7757 5"6242"!72% ,57/5# " ,2"JK$J.K$J4K$JK$ 6J2K%72 "5#6* "557722"7""22656=5#2/$,26722#2%2- ;264!%687242"+FA<=5#2"/<>I+<2>*8" 56=5#2/$ %72,2"%72"56#26-5A "6%722#23,2% 46.25:226775#2%2;26452"$,545642"2"72 " *87%2;26452",276+FA$72, ""5L242,57722#23,2%* MNO C6464"56$5746.2"57772" %&'(2"5"5655A2.!"56#72523 #57*=722$ "75267" ! 5#2&'(2"$,5424"2.!65642" "75:62"*8"2"7$72".9247"$,575642"2 725"75:62"46.222"5!5: % 27!".9247" %45"45"2" *8"%72"7!72242E2 ".246" " C65756"2#57".242,57 "55725"75:62"2"757562% "* PQROSTUVSW 676*7/)+$.!72X"7656?#272=566456#9247 3=)2Y))37")7$6.!72X26Z6567#72 X26[2# =6* \S]S^SS 754"2,22457!2"2.! ;2"a4-5$"!#46725#$Kbcdefghcijklmenmfopcf$ *3+$6*2)$*3))2I3))7$3))0* 1343*3*352$F*X*855$*F*=*[27*$J67557! 5#56#%72"42754;247557!64"5756 "56#= -X?q5?66[C564756%27252"$K &nrmecoshmecopo$*3)7$6*2$*2tui2yt$3))u* 5"27"66#627542"66425#56#72465;2"$K &nrmecoshmecopo$*32/$6*3$*y)/iy2)$3)2* 1Y45*=*526.4$8*372,"$8**[*w6 2"7227*$J5-5"6%727!"# ! "%725"75:62"$Kjklmenmfopcfxmomgesr$* t/$6*3$*330i3tt$3)23* 1/4*826$J727!"#!657" !"5#542"2267$Kyrkopchczpsgh{mgodem mfn$* 30$*[2I[tu$3))7* 1+48*8*[*_$*`*F62"$ *C56#$68**626"$ ldnme{mnrco gfyeczeg opf pc mpspfm$*30$6*y$*3/7i3+u$2u0u* 174v* *v56.2#6* 8* ƒ.2#$jƒ "% $K lnpsgh fzpfmepfz$*t3$*3/ti3/7$2uut* #54"5#642"56#"56#752"4.872% "22!2"2267$K oncfpgfbcdefghci fzpfmepfz$* 2+$6*2$*70IuY$3)2)*

95 '16(& A(9E4153D&"A95695&49G6(1'(969'16(& 9!"# $ )'(43613(54*1+,-./ !9: %&'(9 :;CDEC5RY4616"#1 :QB%45"R45)151'(969'16(& '1 5&4%&7645)E46(S45+T012/3K3J1U2.K.V2/3K 1166(175)1+H.I-J3K.LM9N0-80J72.J!9;< 9) E!1G9 :=(4!9364!1)1565!3)45)%&6( !64!1 9G'(969'16(& 969Z86%1#4 9) E!1G9 +M9N0-80J72.J!9=;59 16"Z G :<C\6CE6 9G6( !64!19G6(175)1'(969'16(& 5%3(451'1E!1!19346&[01\( ]60-2/3JH.I-J3K.LM9N0-80J72.J!9:O59<''<XX><OP 65A^%4"Q6446&9G 9) 6&+ G96(1 9) 4556 :PC\6CE :?#S%9!4`3451!43`64516"Q%446&9G '(969'16(& "^51' )1 9)'(& 45569G139514!64!19G6(175)1 95)!49349!34G3 16( A^% G & :X2A41BF1&B4615FA(9E4153D& "01!34 64$51+45a.b0KU2.N5.8.J2/c0/5J2dI073J1]NNK2/382.J7!9 :C8F53A15"A(1%&(1!''9h4 :OF99%999"2 EEE(99% !64!1g 4959G6E94)46!9 '369G)45)5!9364!1)[39 996(59419%64$ (6'[g g'!1g5 1' (9g5 156+]60-2/3J 6495G9 143 ' ;:F4614)4)4 2J] oi0.-082/pWqN0-260J83K3J14K2J2/3K,-2J/2NK07 1h' !14)46761E46(451'(1+fWWWc-3J73/82.J7 9G61464$51+45,-./012JV7.L850rsJ1fJ80-J382.J3K 42-/I28i0.-9!9:59;'':O>:<:X; 4.JL0-0J/0.L850fWWWWJV2J0-2JV2JT012/2J03J1U2.K.V C1 '1hG9'9%% "#5 6495

96 ÿ ÿª ««ÿÿ ÿ ±² «EFGHGIJK L HMNIJOPQ ÿ89:;<9=:>?9=@ÿa?b<9=@ÿ?c ÿ Ë ÌÍ ÿøïùÿ ÐÑÚÍÖÒÛÓ ÑÒÔÜÐØÿ ÕÖ ± ² ÿ µ ÿ ¹ ºÿ ÿ»¼ ½ ¾¼ ½ ± ¼ ÿ ¹ ¾À Á Á ²²²Â ¹ ± ²  à ¼ Ä ¼ Å µ ÄÆ ÿç È ªÉ «Ê ðéñãåëÿð ÿ ÿ689ÿ 314ÿ ÿ áâãäåæâÿèéêëâìíâãîÿïðñòðñåóâðã íóóòôõõæææöíâãäåæâö ðø ùðëéøúÿûüýþ R>9S=T>ÿUBV@>WX>9YÿZ?<[?<=:>?9 X::[\]]TTT^X>9S=T>^_?` a?@b`;ÿbcde U=<:ÿ8 ÿ ÿ ÿ unvhjpÿu hx v h j xg u h j ÿpÿ fghijkgÿmnopgqrghsÿtuvwuvjxguh rxxwyzzkkk{rghijkg{ u} ÝÔÏÍÜÞÿßàáâ ÀÁÂÃÄÁÅÂÆÇÁÅÈÿÊÇËÄÁÅÈÿÇÌ ÍÆÁÎÅÏÆÿÐËÑÈÆÒÓÆÁÔÿÕÇÄÖÇÄÅÂÆÇÁ ÓÂÂÖ ØØÏÏÏÙÓÆÁÎÅÏÆÙÚÇÛ ÿ7819ÿ ÿ ÿ ÿ ÿ!ÿ" # # >?@ ABC DEEÿÿKGB ÿ7 >C HIE J G C 87 LGMNAJGHB '! ( ÿ)%* +, #ÿ -, -.//(((0,! ( 0 1 OGB?APGÿQIMRGESGBTÿUHNVHNAJGHB SJJVWXXPPPYSGB?APGYCHZ $ % ÿ & ÜÇÈËÛÃÿÝÞßà ñòóôõõö øÿúÿ ûöüýõþÿöò ÿ ö ÿ0 øö ôô1ö ø m } sÿ }~ ÅÆÇÈÉÊÆÿÌÍÎÏÆÐÑÆÇÒÿÓÔÕÖÔÕÉ ÆÔÇ Ñ ÖØÙÙÊÊÊÚÑÆÇÈÉÊÆÚÛÔÜ ³ µ ¹ºÿ¼½¾ ½µµ ½¾ ~upn} ÿ mnopÿrstuvw xuwyz{uÿ }~ ut uwvÿ zu w {{{ uwyz{u ïèðéçûþðÿæç ãä ìèíåéæèçäèäéîåéèçíêÿ ÿ ÿ 7 ÿ!"#$%&'ÿ)$ *+,-./&%,0-$)%' ÿtuvwxyzÿu{ÿ UVWXYZVÿ\]^_V avwbÿcdefdeygvdw aggfhiizzzjavwxyzvjkdl ()*+ÿ -./ :;<8ÿ>?@A8BC89DÿEFGHFG;I8F9 CIIHJKK<<<LC89:;<8LMFN md_]lnÿopqr OFA?NPÿQRST 2 %1 ÿ3456 ÿ ÿ!"#$%!ÿ'()*!+,!"-ÿ./01/0$2!/",221344%%%5,!"#$%!56/7 8/*(79ÿ:;<= ÝÞßàÿ ßâãâäåæç èéêëìíéÿïðñòéóôéêõÿö øù øìúé ê ôúúùûüüíííýôéêëìíéýþ ÿ LHRIZDÿ[\]^ 0 òðÿ1ÿ ÿ 7 ÿ ! " # "9ÿ$%&' ÿ ÿ ª«ÿ ± ²³ ÿµ «¹ ³¹¹ º»» ¼³ ª«¼½ ¾ ± ¾ÀÿÁÂÃÄ 1)$!#2)ÿ345/)'6)$7ÿ8-0+-0#,)-$ 6,,+9::222;6)$!#2);%-< =-/4<&ÿ>?@A BCDEÿ CGHIJKÿEGLMJKNI OPQRSTPÿVWXYPZ[PQ\ÿ]^_`^_SaP^Q [aa`bcctttd[pqrstpde^f g^ywfhÿijkl