Medical Engineering & Physics

Transcription

1 Medical Engineering & Physics 31 (2009) Contents lists available at ScienceDirect Medical Engineering & Physics journal homepage: Ballistocardiogaphic studies with acceleration and electromechanical film sensors J. Alametsä a,c,, A. Värri b, J. Viik a, J. Hyttinen a, A. Palomäki c a Tampere University of Technology, Department of Biomedical Engineering, Tampere, Finland b Tampere University of Technology, Department of Signal Processing, Tampere, Finland c Kanta-Häme Central Hospital, Hämeenlinna, Finland article info abstract Article history: Received 26 February 2008 Received in revised form 29 May 2009 Accepted 21 July 2009 Keywords: EMFi Ballistocardiography Blood pressure Acceleration sensor The purpose of this research is to demonstrate and compare the utilization of electromechanical film (EMFi) and two acceleration sensors, ADXL202 and MXA2500U, for ballistocardiographic (BCG) and pulse transit time (PTT) studies. We have constructed a mobile physiological measurement station including amplifiers and a data collection system to record the previously mentioned signals and an electrocardiogram signal. Various versions of the measuring systems used in BCG studies in the past are also presented and evaluated. We have showed the ability of the EMFi sensor to define the elastic properties of the cardiovascular system and to ensure the functionality of the proposed instrumentation in different physiological loading conditions, before and after exercise and sauna bath. The EMFi sensor provided a BCG signal of good quality in the study of the human heart and function of the cardiovascular system with different measurement configurations. EMFi BCG measurements provided accurate and repeatable results for the different components of the heart cycle. In multiple-channel EMFi measurements, the carotid and limb pulse signals acquired were detailed and distinctive, allowing accurate PTT measurements. Changes in blood pressure were clearly observed and easily determined with EMFi sensor strips in pulse wave velocity (PWV) measurements. In conclusion, the configuration of the constructed device provided reliable measurements of the electrocardiogram, BCG, heart sound, and carotid and ankle pulse wave signals. Attached EMFi sensor strips on the neck and limbs yield completely new applications of the EMFi sensors aside from the conventional seat and supine recordings. Higher sensitivity, ease of utilization, and minimum discomfort of the EMFi sensor compared with acceleration sensors strengthen the status of the EMFi sensor for accurate and reliable BCG and PWV measurements, providing novel evaluation of the elastic properties of the cardiovascular system IPEM. Published by Elsevier Ltd. All rights reserved. 1. Introduction 1.1. Early history of BCG and systems used In this article, findings on the applicability of electromechanical film (EMFi) and acceleration sensors in recording ballistocardiography (BCG) are presented. Using these modern sensor technologies, this method has promising potential. Because the method is not well known, a short review of the method and its history is included. Corresponding author at: Tampere University of Technology, Department of Biomedical Engineering, P.O. Box 553, FIN Tampere, Finland. Tel.: ; fax: addresses: jarmo.alametsa@tut.fi (J. Alametsä), alpo.varri@tut.fi (A. Värri), jari.viik@tut.fi (J. Viik), jari.hyttinen@tut.fi (J. Hyttinen), ari.palomaki@khshp.fi (A. Palomäki). BCG is a recording of the movements of the body caused by shifts in the center of mass of blood in the arterial system and, to a lesser extent, of the heart, caused by cardiac contraction [1,2]. A mass displacement with a certain velocity and acceleration inside the body originates with the heart s action [2]. A ballistocardiograph measures the motions of the body produced by cardiac contraction [3] by recording the resulting force vectors in the long axis of the body [4]. In 1877, Gordon recorded the first published results about ballistic displacements of the body by placing a subject on a bed suspended by ropes from the ceiling (pendulum bed) and obtained a record of its motion (displacement) synchronously with the patient s heart beat [3]. This recording device was impractical, as the amplitude of the movement of the ballistocardiographic bed caused by respiration was many times higher than that caused by circulatory events [2]. As the respiration of test subjects was suppressed during measurement, it was difficult to control the level of suppression [1] /$ see front matter 2009 IPEM. Published by Elsevier Ltd. All rights reserved. doi: /j.medengphy

2 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Modern BCG was born in 1939, when Isaac Starr presented a ballistocardiograph that was a modification of the earlier pendulum bed [3]. As in the older model, the bed was suspended with cables attached to the ceiling, the lateral movement was constrained, and the displacement of the bed was recorded. To minimize the effect of respiration, Starr constructed a high-frequency bed in which the motion of the bed was opposed by a stiff spring to permit motion only in the longitudinal direction, thereby making the entire measuring system into a huge strain gauge [5]. By measuring the displacement of the system, the force (F = ma) and the acceleration were obtained. Upon displacement of the system, the movement of the body s center of gravity in space was recorded [6]. Inthe high-frequency BCG [7], the natural frequency of the measurement device was greater than 10 Hz. However, the high-frequency bed was so strongly coupled to the surroundings that it did not perform the same movement as the body lying on it [2]. Another disadvantage of the high-frequency measurement system was the natural frequency of the system, which was within the frequency range of the BCG (from DC to 40 Hz), causing distortion in the BCG waveform [8]. The third disadvantage was the need for firmly coupling the tested subject, which was not usually used, with the heavy bed, resulting in additional distortions of the BCG signal. The fourth disadvantage was the strong coupling of the heavy instrument to the ground or the building. Inevitably, vibration of the building interfered with the tracing [1]. Finally, it became obvious, that in the Starr high-frequency system, the Nickerson low-frequency system, and the Dock direct body system, information regarding circulatory function was seriously distorted, mainly due to the way in which the body was supported [7]. In these approaches, the body tended to oscillate on its dorsal spring with respect to its support, which initiated a large resonance peak occurring at the natural frequency of the body with marked attenuation at higher frequencies [2,9]. Therefore, the bed supporting the subject had to be extremely light, because heavy beds created serious distortions. Additionally, the coupling to the ground should be minimized (i.e., weakly bound to surroundings) and the frequency of the system should be preferably less than 0.3 Hz [1]. In the ultra-low-frequency BCG, the subject is free to move and acceleration is recorded to measure the force [10]. The simplest ultra-low-frequency BCG bed used was a pendulum bed composed of an aluminium-canvas bed suspended from cables attached to the ceiling. Hip and shoulder straps were used to increase coupling between the subject and the bed. An accelerometer attached to the bed recorded body movements [1]. The pendulum bed was complicated to use and was sensitive to building vibrations. In another system, the bed floated on air bearings with 0.14 Hz of natural frequency. The disadvantages of the air-bearing beds were the noticeable building vibrations and the smoothness of the patient contact surface, resulting in poor subject bed coupling and distortion [1]. The modern transducer used in BCG recordings was based on the piezoresistive principle, in which the crystal changed its resistance when distorted. In earlier studies, the transducers used were based on coils and piezoresistive crystals and were attached directly to the bed [1]. The loose coupling of the subject to the bed and rigid supporting structures of the bed produced results with variable sensitivity. Different types of apparatus used in obtaining ballistocardiograms varied in their frequency response [7]. However, a recent technological development provides new sensors suitable for BCG recordings, whose technical features are superior compared with earlier BCG systems Definition of ballistocardiographic waveforms When the heart pumps blood from ventricles to the pulmonary arteries and ascending aorta and through the aortic arch to the Fig. 1. Signals were recorded in the sitting position with suppressed respiration. (a) Blood pressure 147/93, heart rate 67. Signals from top to bottom: signal from the carotid artery (carotid pulse) recorded with an electromechanical film (EMFi) sensor strip, signal from the EMFi sensor on a chair, electrocardiogram (ECG), and phonocardiography (PCG). (b) Typical, normal BCG waveform with the letters used to identify the component parts as reported in the literature and measured with traditional methods [13]. The arrow indicates the beginning of electrical ventricular systole. peripheral circulation, the recoil in the opposite direction is applied to the body, and its force and direction change according to the cardiac cycle. BCG waveforms (Fig. 1) have been divided into three groups: pre-ejection (FGH), ejection (IJK), and the diastolic portion of the heart cycle (LMN) [1]. Pre-ejection waves consist of venous return to the heart, atrial filling, and contraction (H; head-ward deflection). The foot-ward deflection, I, reflects the rapid acceleration of blood in the ascending aorta and pulmonary arteries around the aortic arch and into the carotid arteries. The ejection phase J- wave describes the acceleration of the blood in the descending and abdominal aorta and deceleration of blood in the ascending aorta. The peak of the J-wave corresponds to the end of rapid ejection of both ventricles [11]. I J amplitude reflects the force of contraction of the left ventricle and I J period reflects contractility. The K and L waves reflect the deceleration and cessation of blood flow and the closing of the aortic valve [1]. Diastolic waves (KL and MN) reflect the state of peripheral circulation. In addition, the influence of arterial wall stiffness and peripheral resistance has greater influence on the diastolic waves [12]. In Fig. 1b, a typical, normal BCG waveform

3 1156 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Fig. 2. The ADXL202 acceleration sensor: (a) wiring diagram including amplifier and filtering sections, and (b) circuit board. The light emitting diodes (LEDs) and integrated circuits (ICs) (related to the movement indicator) depicted in (b) were not used for this study, but give a size comparison reference for the acceleration sensor. The MXA2500U acceleration sensor (c) wiring diagram, including amplifier and filtering sections, and (d) the circuit board (left) and the acceleration sensor (right). is shown as reported in the literature and measured with traditional methods [13] Carotid and ankle pulse waveforms The carotid pulse rises sharply with the ejection of blood from the left ventricle to the ascending aorta, reaching a peak called the percussion wave (P) (Fig. 1a). The subsequent secondary wave is called the tidal wave (T), caused by a reflected pulse returning from the upper body. The dicrotic notch (D) is caused by a closure of the aortic valve and can be followed by a dicrotic wave (DW), which is due to the reflected pulse from the lower body [14] Objective of the study The objective of this study was to develop a new measurement system to record improved BCG and pulse wave data with different EMFi measurement configurations. The primary objective was divided into two parts. First, the feasibility of using two acceleration sensors and an EMFi sensor was evaluated qualitatively in BCG and pulse wave velocity (PWV) research. Second, the utilization of the EMFi sensor and the functionality of the proposed instrumentation were evaluated under different physical conditions. 2. Materials and methods In previous BCG studies, an acceleration sensor attached to the bed recorded body movements. In these structures, inadequate coupling of the subject to the bed, among other structural defects, caused artefacts and nonlinearities in the measured signal. In this study, we placed sensors directly on the skin of the subject (except the seat EMFi sensor) to ensure as close a link as possible between the changes in blood inertia and the sensor Acceleration sensors One method to noninvasively measure the ballistic forces generated by the heart is to measure the tiny displacements of the skin by attaching an acceleration sensor directly onto the skin of a measured person. The method for coupling the sensor to the skin, the sensor properties (sensitivity, dimensionality, noise rejection, weight and supporting structures), and wiring solutions has their own effect on the measured signals ADXL202 acceleration sensor The ADXL202 acceleration sensor is a low-cost, low-power, twoaxis accelerometer (Fig. 3) with a measurement range of ±2g and the sensitivity of the analog output is 312 mv/g [15]. The typical background noise level is 500 g/ Hz. With a 10-Hz bandwidth, the noise floor is mg and the rms noise is 1.9 mg. More details of the properties of this sensor and the MXA2500U sensors can be found in their respective data sheets [15,16]. The ADXL202 acceleration sensor can measure both dynamic acceleration (shock/vibration) and static acceleration (gravity). The duty cycles of the digital output (ratio of pulse-width to period) are proportional to the acceleration in each of the two sensitive axes. The sensor and three resistors were attached to a printed circuit board slightly larger than the sensor chip itself. The movement of the sensor in the X and Y directions was measured with the digital outputs, X OUT and Y OUT. The period of the duty cycle modulators repetition rate is 1 khz or 1 ms. The digital signal was filtered with a passive RC filter of f C = 12.5 Hz. The low-pass filter converted the pulse-width modulated digital signals to an analog form. These output signals were amplified with an AD623 instrumentation amplifier and were filtered with a high-pass filter having a first-order RC circuit with the f C = Hz (Fig. 2a) to remove the DC component from the measured signal. Finally, a third-order, anti-aliasing filter with f C = 240 Hz prepared the signal for A/D con-

4 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Fig. 3. The two MXA2500U acceleration sensors are the outermost sensors; the sensor in the middle is the ADXL 202 acceleration sensor. The mounting positions for the ADXL 202 sensor and the MXA2500U sensors were similar. In the picture, the X direction is upwards. The physical sizes of the different sensors have their own influence on measured signals. version (usually with a 500-Hz sampling frequency). Fig. 2b shows a photograph of the sensor and the analog amplifier/filter connected to it. Fig. 4. Two electromechanical film (EMFi) sensor strips used in the pulse wave recordings from the limbs and on the neck on top of the larger EMFi sensor. The smaller strips are 11.4 cm 2.1 cm (left) and 15 cm 2 cm (right). The larger EMFi sensor used in seat recordings is 42 cm 36 cm MXA2500U acceleration sensor The MXA2500U acceleration sensor (Figs. 2c and 3) is a lowpower, low-cost, two-axis accelerometer with a measurement range of ±1 g and a sensitivity of 500 mv/g [16]. The MXA2500U acceleration sensor uses a bubble of hot gas as the proof mass that yields two benefits: high shock resistance and low noise. The MXA2500U acceleration sensor provides an absolute analog output. The typical background noise level is 0.2mg/ Hz allowing signals below 1 mg to be resolved at a 1-Hz bandwidth [16] Sensor placement The acceleration sensors (Fig. 3) were used to measure the movements of the chest and the recoil of the heart. Sensors were attached in the electrocardiogram (ECG) electrode position, V4, and also the MXA2500U sensor on the neck near the carotid artery. The sensors were attached directly to the skin with fast-drying glue. As a whole, the sensor consisted of the acceleration sensor, the preamplifier, and an analog filtering unit (Fig. 2c). The acceleration sensors were attached in such a manner that the X direction was forward from the chest and the Y direction was down from the chest EMFi sensors The EMFi is a type of capacitive film-based sensor made of a thin, biaxially oriented polypropylene film coated with metal electrodes (Fig. 4) [17]. The thickness of the film for sensor applications is typically m, the frequency response is inherently flat to greater than 20 khz, and the sensitivity value is pc/n, which exceeds the values of other piezoelectric polymers. A change in the pressure or force caused by introducing mechanical compression into the film changes the spatial distribution of charges within the material. As a result, a mirror charge, which is proportional to the force or pressure, is induced at the surface electrodes. This charge can be measured as a current or voltage signal by a charge amplifier. The signal from the EMFi sensor is caused by the movement of the charged polymer layers with respect to the other layers and is not of piezoelectric origin [17]. The EMFi sensor can convert mechanical force into proportional electrical energy and vice versa. Therefore, the EMFi sensor acts as a sensitive movement sensor suitable for BCG recordings. The EMFi sensor has also been used in sleep research by the author to evaluate the occlusion rate of the pharyngeal structures (snoring) observed as spiking in EMFi sensor sleep recordings [18]. In an attempt to obtain a high-quality BCG signal with a lightweight, flexible, and extremely sensitive sensor (high- sensitivity EMFi; sensitivity 703 pc/n), the EMFi sensor was included in the BCG studies. In seat recordings, the size of the EMFi sensor used was 42 cm 36 cm (Fig. 4). For the carotid and ankle pulse wave measurements, smaller EMFi sensors are required and the sizes of the EMFi strip sensors were 11.4 cm 2.1 cm and 15 cm 2 cm, respectively Biosignal measurement unit A mobile physiological signal measurement station (Fig. 5a) and its current configurations (Fig. 5b and c) were used as devices that enable the recording of BCG, carotid, and limb pulse signals with the EMFi sensor and the acceleration sensors [19,20]. The first version [19] included four ECG channels: one for heart sound and three for EMFi sensor channels (Fig. 5a). The second version [20] (Fig. 5b) includes 12 channels for the ECG (12-lead ECG measurements), 1 channel for heart sounds, and 3 channels for EMFi sensors. The entire device is battery operated. The gains of the ECG channels can be set separately, from 40 to 72 db, allowing also the electromyography and electroencephalography to be recorded. All channels are isolated with optocouplers and filtered with an active, low-pass filtering section (eighth-order Butterworth filtering, having a cut-off frequency of 256 Hz; thereby, the sampling frequency usually used is 500 Hz). In addition to improved patient safety, the usage of optocouplers and active filtering improves the quality of the measured signals. The construction of the measurement device was planned to generate the most realistic signal and avoid cross-talk between channels. For this reason, with every active component, its own power delivery was implemented. With multiple switches, it is possible to switch on the needed channels in the device. This reduces interference and increases battery duration. In Fig. 6 a block diagram of the second biosignal measurement unit is presented. The comprehensive recording setup for monitoring the limbs and the neck with the EMFi sensors (Fig. 7) made it necessary to construct a new biosignal amplifier for arterial elasticity measurements (Fig. 5c). The new amplifier had an added number of EMFi channels with the same inner configuration as in the previous devices. By using smaller batteries, the size and the weight of the device decreased without the need for sacrificing the safety issues used in the older devices. The second device was authorised for clinical use in Tampere University Hospital. By keeping the inner structure of the measurement device the same, the clinical authorization procedure remained simple also for the latest version.

5 1158 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Fig. 5. (a) The main unit of the first mobile physiological signal measurement station: electromechanical film (EMFi) sensors were used in the supine recordings, phonocardiography (PCG) electret microphone, and cup electrodes were used in the primary ballistocardiography (BCG) studies. (b) The main unit of the second version of the physiological measurement station. The device is battery operated, thereby improving patient safety and the quality of the measured signals. (c) The latest version of the mobile physiological signal measurement station, which includes 1 electrocardiogram, 1 PCG, and 7 EMFi channels for limb and neck recordings. The seat EMFi sensor has a lilac-coloured shelter padding and its size was 29 cm 30 cm Measurements ECG, BCG, carotid, ankle pulse, and, in some recordings, heart sound signals were measured. BCG measurements were performed with three sensors (ADXL202, MXA2500U, and EMFi). Carotid pulse was measured with MXA2500U and EMFi strip sensors. The MXA2500U acceleration sensor was used to make qualitative comparisons with the seat EMFi sensor signal. The blood pressure (BP) and the heart rate were measured just before obtaining the measurements with the Omron M5-I blood pressure monitor device from the brachium of the right arm of the subject. All measurements were conducted on the same person (41-year-old male, height cm, weight 98 kg). The total measurement lasted about 3 min, including suppressed respiration in the beginning of the measurement, which lasted about s. To acquire information about the stiffness of the arteries and the response of the heart under changing conditions, EMFi sensor strips were attached on the neck near the carotid artery and on the ankle near the femoral artery (Fig. 7) along with the seat BCG recording. Temporal values for the pulse transit times (PTT) were calculated from the peak of the pulse waves (carotid and ankle pulse waves with EMFi sensor strips) or from the peaks of the BCG waves (with seat EMFi sensor). For the MXA2500U PTT measurements, acceleration sensors were attached on the chest and on the neck (Fig. 7; positions 1 and 2). The adaptation of the cardiovascular system in the case of hypertensive values was studied via physical exercise and sauna bath in sitting position. The decline of the resistance of the circulatory system was induced first by physical exercise and then with a sauna bath in which vasodilation of the veins was generated by the heat of the sauna bath (temperature 80 C). Bathing in the sauna was used to increase the heart rate without physical exercise. The aim of the physical exercise (duration 1.5 h) was to increase the pulse level to the sweating level (50 60% from the maximal heart rate). The sauna and exercise experiments were separated more than 1-month time interval. The effect of suppressed respiration on the BCG signal was also studied in sauna and exercise recordings Processing the data The sampling frequency was 500 Hz. Signals were first-band pass filtered using Matlab software ( Hz FIR, 700 taps, time delay corrected) and down sampled into 100 Hz. The analysis was conducted with an analysis window length of 1.08 s, which moved in 0.5 s increments. The R wave of the ECG was used for reference for all detected waves and the index of the R point was detected first by differentiating (2 points), squaring and integrating (5 points), and by taking the maximum from the ECG signal. The H slope from Fig. 6. A block diagram of the second biosignal measurement unit consisting of the electrocardiogram amplifier (AMP02) section of the system isolated with an optocoupler (IL300). From the second version of the biosignal measurement devices, all of the channels were isolated with optocouplers for safety reasons and improved noise rejection ability.

6 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Fig. 8. Signals recorded in a sitting position with the ADXL 202 acceleration sensor, suppressed respiration, and 40-Hz low-pass filtering. Signals from top to bottom: signal from the Y direction (reversed to show similarity with X direction) and from the X direction (forward from the chest) measured with ADXL202 acceleration sensor, electrocardiogram, and phonocardiography. trum was calculated from the raw signal (1024 length fast Fourier transform). The changes in ballistic complexes in the time and spectral domain were studied. In all figures, where waveforms of the recordings are presented, the vertical scale of the figures is arbitrary as the signals are not calibrated. Comparison between the measurements before and after exercise as well as before and after sauna bath was made using Student s paired samples t-test. All statistical analyses were performed using the SPSS 13.0 program (SPSS Inc.). 3. Results 3.1. Acceleration sensor measurements Fig. 7. A human torso showing the positions of the sensors. Labels 1 (MXA2500U) and 2 (ADXL202 and MXA2500U) depict the positions of the acceleration sensors. In acceleration sensor measurements, the X direction was oriented anteriorly and the Y direction was oriented downward. Labels 3, 4, and 5 (dorsalis pedis pulse from the ankle) illustrate the positions of the electromechanical film (EMFi) strip sensors; label 6 illustrates the larger seat EMFi sensor. the BCG was detected using the local maximum method, and the I slope was detected using the local minimum with the index of the H point as a starting point. Other slopes were detected in the same way. Based on the detected points, temporal (T HJ, T RI ) and amplitude (A HI, A IJ, A JK ) values from the systolic BCG components were acquired. Temporal (T RP, T RD, T RV ) and amplitude (A PD, A UV ) pulse wave values were acquired from the maximum value of the carotid and ankle pulse signals. PWV was calculated from the T RV value and by using the distance from the jugulum point to the ankle EMFi strip location (Fig. 7; strip 4). For MXA2500U PTT acceleration sensor recordings, temporal (T RA, T AC, T RD ) and amplitude (A AB, A BC, A DE ) values were acquired from the chest BCG and carotid pulse signals. The amplitude spec- The ADXL202 acceleration sensor was able to trace the movements of the chest caused by the pumping action of the heart. The tracings (Fig. 8) resemble closely the typical BCG complexes, especially in the X direction taken from the electrode position V4 (Fig. 7; position 2) according to the 12-lead ECG definition. Strong fluctuation of the X direction was seen in the same recording because of the normal respiration. The MXA2500U acceleration sensors (Fig. 7; positions 1 and 2) were used to measure the BCG from the chest (Fig. 9a; X and Y directions, Fig. 9b; X direction), and the carotid pulse signal from the neck (Fig. 9b). When the MXA2500U acceleration sensor trace was compared visually to the simultaneous trace from the EMFi sensor (Fig. 9b), the X direction signal from the chest resembled closely the BCG signal from the seat EMFi sensor. The small time delay between measured signal components observed in (Fig. 9b) is due to the different distances of the sensors in proportion to the heart location. Because the MXA2500U measurement site is situated near the heart on the chest, more time is needed for the pulse wave to travel into the seat EMFi sensor. The elevated BP values dropped because of exercise. Amplitudes from the chest (A AB, A BC ), neck (A DE ), seat BCG decreased after exercise with normal respiration (Table 2). According to the t-test, all amplitude values (except A BC in the suppressed part of the respiration) were statistically significant, with both suppressed and normal respiration parts showing dissimilarities between values before and after exercise. In time domain

7 1160 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Table 1 Mean values of amplitudes (EMFi BCG); A HI, A IJ, A JK, (MXA2500U); A AB, A BC and A DE (Fig. 9b) ± SD from the section of suppressed respiration and normal respiration before (BP 154/103, heart rate 71) and after (BP 131/84, heart rate 87) physical exercise. Correlation coefficients (paired samples t-test), where the correlation is significant at the 0.05 level (2-tailed) is mentioned in the p value column. NS means not significant. and after supp. exercise) Length 36 s 36 s 136 s 138 s and after normal exercise) A HI 5189 ± ± 493 < ± ± 638 <0.001 A IJ 4485 ± ± 351 < ± ± 772 <0.001 A JK 2232 ± ± 748 < ± ± 619 <0.001 A AB 154 ± ± 20 < ± ± 30 <0.001 A BC 205 ± ± 22 NS (0.056) 214 ± ± 42 <0.001 A DE 250 ± ± 46 < ± ± 41 <0.001 Table 2 Mean PTT values (EMFi BCG); T HJ, T RI, (MXA2500U); T RA, T AC and T RD (Fig. 9b) ± SD from the chair BCG, and the chest and neck MXA2500U signals before (BP 154/103, heart rate 71) and after (BP 131/84, heart rate 87) physical exercise. and after supp. exercise) Length 36 s 36 s 136 s 138 s and after normal exercise) T HJ (s) 0.16 ± ± 0.03 < ± ± 0.05 NS T RI (s) 0.20 ± ± 0.01 < ± ± 0.05 NS T RA (s) 0.12 ± ± 0.03 NS 0.10 ± ± 0.05 <0.001 T AC (s) 0.10 ± ± 0.02 NS 0.12 ± ± T RD (s) 0.13 ± ± 0.01 < ± ± 0.02 NS values, statistically significant results were seen in the T RD value in the suppressed part of the study and the T RA and T AC values in the normal part of the study showing differences in the signals before and after exercise (Table 1). Similarities in the T RA and T AC signal mean temporal values are explained by the suppressed respiration EMFi sensor measurements Physical exercise and BCG The experimental results of the BCG before and after exercise in the case of mildly hypertensive values are shown in Fig. 10a and b and Tables 3 and 4. All the measurements were conducted in a sitting position. In this recording, the amplitudes of the BCG complexes were not altered by exercise, except the A UV value from the left ankle, which increased (Table 3). The ejection time of the heart increased after exercise (normal respiration), except the carotid time, T RD, which decreased. PWV values from the left ankle dropped because of physical exercise, observed also as decreased BP values. As observed upon inhaling, suppressed respiration increased the amplitudes of the systolic BCG components when compared with normal respiration. The BP values dropped because of physical exercise. Changes can be seen in the BCG complexes over time (Fig. 10b) and spectral domain (Fig. 11). I J amplitudes of the BCG decreased in the normal BP condition (post-exercise hypotension). In the spectral domain, the fundamental frequency of the heart rate is about 1.1 Hz. According to the t-test, all amplitude values (except A HI in the normal part of the respiration) and temporal values were statistically significant, with both suppressed and normal respiration parts showing dissimilarities between values before and after exercise (Tables 3 and 4). Similarities of the amplitude A HI may indicate a stable pre-ejection of the heart in the part of normal respiration Sauna bath and BCG In sauna measurements, elevated BP values were lower after the sauna bath and the A IJ amplitudes of the BCG almost doubled when compared with hypertensive values before the sauna bath (Table 5). The same effect was also seen in the amplitude of the A PD carotid pulse signal. Interestingly, the PTTs were longer before the sauna bath than after it in this case (Table 6). The amplitude rise of the main systolic complexes can be seen in Fig. 12a and b. Suppressed respiration increased the amplitudes of the systolic BCG components as observed in the physical exercise study. PWV values from the left ankle seemed to increase due to sauna bath (Table 6; normal respiration) denoting increased BP in aorta. According to the t-test, all amplitudes (except ankle pulse amplitude A UV in the normal part of the respiration) and temporal values (except the carotid temporal component T RD in the suppressed part of the respiration) were statistically significant, with both suppressed and normal respiration parts showing dissimilarities between values before and after sauna bath. Similarities of the ankle pulse amplitude A UV may indicate stability after sauna bath too. The stability in the value of the carotid temporal component T RD (suppressed part of the respiration) may denote a stable ejection time seen in the carotid pulse Latest research measurements As a demonstration of the final measurement system, the limb pulse tracings were introduced with the carotid artery and seat BCG signals to acquire information about the stiffness of the entire arterial tree (Fig. 14). All time domain pictures, except Fig. 14, are from the suppressed part of the respiration. 4. Discussion This paper describes the development of BCG and pulse wave measurements with different kinds of EMFi and acceleration sensors. We also present instrumentation that overcomes the limitations of previous BCG devices and assist in obtaining new and improved BCG data. Results of EMFi sensor measurements in evaluation of cardiovascular elasticity are promising and invite more studies in the sub-areas of this research. The first version of the mobile physiological signal measurement station and its configurations, presented in this research, was able to amplify good quality signals for recording because of their inner structure to reject interferences and possible coupling between adjacent channels. When the short- and long-term alterations of BCG, times spanning from a couple of days to 1 year, were studied [21], the chair BCG and carotid pulse signals were recorded with the EMFi sensors. The earlier study showed that the time domain properties of the BCG and carotid pulse signals remained stable

8 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Table 3 Mean values of amplitudes A HI, A IJ, A JK, A PD and A UV (Fig. 10) ± SD from the section of suppressed respiration and normal respiration before (BP 162/102, heart rate 68) and after (BP 136/90, heart rate 81) physical exercise. and after supp. exercise) Length 42 s 45 s 120 s 134 s and after normal exercise) A HI 5962 ± ± 388 < ± ± 723 NS A IJ 5025 ± ± 528 < ± ± 1032 <0.001 A JK 2907 ± ± 420 < ± ± 647 <0.001 A PD 618 ± ± 80 < ± ± 90 <0.001 A UV 579 ± ± 114 < ± ± 98 <0.001 Table 4 Mean PTT values (s) T HJ, T RI, T RP, T RD and T RV (Fig. 10) ± SD from the chair BCG, the neck carotid pulse and the ankle signal before (BP 162/102, heart rate 68) and after (BP 136/81, heart rate 81) physical exercise. PWV values (m/s) from the left ankle pulse wave are mentioned in parentheses. and after supp. exercise) Length 42 s 45 s 120 s 134 s and after normal exercise) T HJ (s) 0.16 ± ± 0.01 < ± ± 0.02 <0.001 T RI (s) 0.21 ± ± 0.01 < ± ± 0.01 <0.001 T RP (s) 0.15 ± ± 0.00 NS 0.14 ± ± 0.00 <0.001 T RD (s) 0.36 ± ± 0.01 < ± ± 0.01 <0.001 T RV (s) 0.27 ± 0.01 (4.69) 0.30 ± 0.01 (4.22) < ± 0.01 (4.53) 0.30 ± 0.01 (4.09) <0.001 Table 5 Mean values of amplitudes A HI, A IJ, A JK, A PD and A UV (Fig. 12) ± SD during suppressed respiration and normal respiration before (BP 162/98, heart rate 65) and after (BP 141/80, heart rate 80) the sauna bath. (before sauna) (after sauna) and after supp. sauna) (before sauna) Length 28 s 28 s 120 s 140 s (after sauna) and after sauna) A HI 5912 ± ± 501 < ± ± 1579 <0.001 A IJ 5410 ± ± 402 < ± ± 2142 <0.001 A JK 2890 ± ± 614 < ± ± 1401 <0.001 A PD 737 ± ± 140 < ± ± 244 <0.001 A UV 1118 ± ± 383 < ± ± 218 NS for the same subject showing that the BCG is characteristic and reproducible. Smaller variations in the contour of BCG waves were probably as a result of changed vascular resistance due to altered blood pressure values. Noninvasive methodologies for the assessment of arterial stiffness include (a) PWV measurements, (b) relating changes in artery diameter to distending pressure, and (c) assessing arterial pressure waveforms [22]. Arterial pulse waves were detected using pressure-sensitive transducers, Doppler ultrasound, and applanation tonometry [22]. In recent studies, limb pulse tracings, along with carotid artery and seat BCG signals, were included to acquire information about the stiffness of the entire arterial tree. These studies may clarify the progression of atherosclerosis by analysing the BCG signal and the propagation speed of the pulse signals in large arteries. Because the BCG is compiled mainly from the contributions of the various larger arteries, their mutual positions and their elastic properties [2] are the reasons the peaks appearing in the BCG may yield more information about the state of human circulation Acceleration sensors Both acceleration and EMFi sensors exhibited comparable results in PTT measurements. The waveform acquired from the acceleration sensors (Figs. 8 and 9) corresponds to a large extent to the BCG waveform measured with the EMFi sensor (Figs. 10, 12 and 14). According to the measurements, the MXA2500U sensor seems to be more sensitive than the ADXL202 sensor. This was expected, because the sensitivity of the MXA2500U sensor was better than the ADXL202 sensor based on the manufacturer s datasheets. Additionally, the weight of the sensor and its electronics has an influence on the recorded signal originating from inertia changes in the body (Fig. 3). The acceleration sensors provided a good estimate of the heart rate calculation in laboratory conditions. When the amplitudes of the acquired signals were compared, the tiny MXA2500U sensor chest signal was inferior in amplitude to the seat EMFi sensor (Fig. 9a). This was observed also when the amplitude values from the neck and chest signals were compared (Table 1). The same relationship was observed, when Table 6 Mean of PTT values T HJ, T RI, T RP, T RD and T RV (Fig. 12) ± SD from the BCG, the neck carotid pulse and the ankle signal before (BP 162/98, heart rate 65) and after (BP 141/80, heart rate 80) the sauna bath. PWV values (m/s) from the left ankle pulse wave are mentioned in parentheses. (before sauna) (after sauna) and after supp. sauna) (before sauna) Length 28 s 28 s 120 s 140 s (after sauna) and after sauna) T HJ (s) 0.16 ± ± 0.01 < ± ± 0.02 <0.001 T RI (s) 0.20 ± ± 0.01 < ± ± 0.01 <0.001 T RP (s) 0.14 ± ± 0.01 < ± ± 0.01 <0.001 T RD (s) 0.37 ± ± 0.03 NS 0.36 ± ± 0.02 <0.001 T RV (s) 0.27 ± 0.01 (4.69) 0.27 ± 0.01 (4.69) < ± 0.01 (4.22) 0.28 ± 0.02 (4.53) <0.001

9 1162 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Fig. 9. (a) Comparison of ballistocardiography (BCG) signals recorded with MXA2500U acceleration sensor and electromechanical film (EMFi) sensor. Signals from top to bottom: MXA2500U acceleration sensor on the chest (chest X and Y), chair BCG measured with EMFi sensor, and electrocardiogram (ECG). To clarify magnitude differences of the signal traces, no multiplication was used in the plot. (b) Signals recorded in a sitting position with two MXA2500U acceleration sensors and EMFi sensor after exercise (blood pressure 131/84, heart rate 87). Signals from top to bottom: MXA2500U acceleration sensor on the neck and on the chest (Fig. 7; positions 1 and 2), chair BCG measured with EMFi sensor, ECG, and PCG. the MXA2500U sensor neck signal amplitude (Fig. 9b) was compared with the neck signal amplitude obtained with the EMFi sensor (Fig. 10b). Moreover, the relatively high standard deviation values in the temporal values detected from the MXA2500U sensor neck and chest signals (Table 2), compared with signals acquired with EMFi sensors (Tables 4 and 6), may presume that the sensitivity to noise and artefacts is much greater with acceleration sensors in human biosignal measurements. When recording tiny movements of the body (BCG signal), other sensor than an acceleration sensor might be a more feasible choice. For this reason, the EMFi sensor was chosen for our BCG studies. Measurements with acceleration sensors might be meaningful in applications in which adaptation from the tiny BCG movements to larger motion detection is important. One possible use would Fig. 10. (a) Signals from top to bottom: electromechanical film (EMFi) sensor on the ankle, on the neck near carotid artery (Fig. 7; positions 4 and 1), beneath the subject on the chair and the electrocardiogram; hypertensive blood pressure (BP) values (162/102, heart rate 68) before exercise; suppressed respiration; (b) Normal BP values (136/90, heart rate 81) after exercise; suppressed respiration. Note the changes in the amplitudes and in the diastolic components of the ballistocardiography (BCG) signal. Scale is similar in both pictures. be smart clothing technology. The usage of acceleration sensors in BCG measurements is possible also by attaching the sensors into more rigid structures, such as beds and chairs, but at the cost of sensitivity. As in past studies [1], the need for strong coupling of the subject and the measurement bed would be difficult to accomplish without causing discomfort to the subject. Another technique is to attach the acceleration sensor directly on the skin of the subject with glue to ensure the necessary reliability in BCG measurements EMFi sensor According to the results of this research, the EMFi sensor appears to be an optimal choice for BCG measurements. In the past, with the old and cumbersome BCG recording devices, high-frequency information of the BCG signal was not recorded because of the physical constraints of the measurement devices. A stiff recording system

10 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Fig. 11. Amplitude spectrum of ballistocardiography (BCG) before (solid line) physical exercise (blood pressure 162/102, heart rate 68) and after (dashed line) physical exercise (blood pressure 136/90, heart rate 81) calculated with 1024 point fast Fourier transform (FFT). The BCG recording was performed 1 h after the physical exercise. The electromechanical film (EMFi) sensor was beneath the person on the chair.. The heart rate frequency is about 1.1 Hz. The frequency area from 4 to 10 Hz correspond the systolic components of the BCG signal. The lowfrequency components from the diastolic part of the BCG signal have been merged to the harmonics of the heart rate spectral component. damped the high-frequency information of the measured signals. Because the frequency response of the EMFi sensor ranges from near zero to hundreds of khz, all high-frequency information from the frequency range (0 40 Hz) of the BCG will be recorded with a suitable preamplifier. Compared with the acceleration sensors, the EMFi sensor strips were easier to attach directly to the skin using tape, thereby causing minor discomfort to the subject. Small deformations of the strips, as a function of the pulse wave in the blood vessels, generated an observable signal in the sensor output. The flat and thin EMFi sensor was the most feasible choice when tested in the chair and night-time supine measurements. The type of sensor has an effect on the amplification required for the recorded signal. With EMFi sensors, the charge amplifier is most frequently used, because it supports all sizes of the EMFi sensors (Fig. 4). In general, the larger-sized EMFi sensor, or seat sensor, produces a greater signal than the smaller strip sensor. Also, the coupling of the subject with the seat EMFi sensor is better because of the weight of the subject. With acceleration sensors, tiny movements of the skin from heart action were greatly amplified, leading to possible noise amplification. EMFi sensors have also proved to be good choices for noninvasive BCGs and pulse detection because of their excellent sensitivity, flatness, lightness, and formability [20]. In an earlier study, we studied the I wave of the BCG with an EMFi sensor and the R wave from the ECG in pulse detection [20]. Both methods provided the same mean heart rate value, showing the potential of EMFi in heart rate detection. When compared with piezoelectric materials, the EMFi sensor material is cheaper in many applications, more durable, and has a wider frequency response. Moreover, in human pulse wave measurements (elasticity of the veins), the use of the piezoelectric sensors is inconvenient because of the necessity for accurate placement of the sensor in proportion to the position of the measured artery to obtain a high-quality signal. With the EMFi sensor, such exact placement of the sensor in proportion to the location of the artery is not needed, thus speeding up the measurement and ensuring that high-quality signals are obtained without the need of highly skilled technicians. The ease in placement and use makes it possible to obtain high-quality signals Fig. 12. (a) From top to bottom: ankle pulse signal, carotid pulse, ballistocardiography (BCG), and electrocardiogram (ECG). Measurement was made with hypertensive blood pressure (BP) values (162/98, heart rate 65) before the sauna bath; suppressed respiration. (b) From top to bottom: ankle pulse signal, carotid pulse, seat BCG, and ECG. The measurement was made after the sauna bath with normal BP values (141/80, heart rate 80). The size of the electromechanical film (EMFi) strips was 15 cm 2 cm and the size of the seat EMFi sensor was 42 cm 36 cm. Typical pressure pulse waveforms in carotid pulse and ankle pulse signal can be seen. Suppressed respiration. Scale is similar in both pictures. and renders EMFi sensors suitable also for mass screening of the population Exercise and sauna measurements The effects of a sauna bath and exercise on aortic compliance and the adaptation of the cardiovascular system via BP and PWV changes were studied in experimental recordings with the EMFi sensor. With both methods, elevated BP values dropped and the internal resistance of the vascular system appeared to decline, noted as elevated amplitude values of the systolic complexes of the BCG signal, particularly in the sauna recordings. After a sauna bath, elevated BP values were lowered and the A IJ amplitudes of the BCG and the A PD carotid pulse signal almost doubled when

11 1164 J. Alametsä et al. / Medical Engineering & Physics 31 (2009) Fig. 13. Amplitude spectrum of ballistocardiography (BCG) before (solid line) the sauna (blood pressure 162/98, heart rate 65) and BCG after (dashed line) the sauna (blood pressure 141/80, heart rate 80) calculated with 1024 point fast Fourier transform (FFT). The BCG recording was performed 30 min after the sauna bath. The electromechanical film (EMFi) sensor was beneath the subject on the chair; suppressed respiration. compared with hypertensive values before the sauna bath (Fig. 12, Table 5). Also, the highest amplitudes in the spectrum (Fig. 13) after the sauna bath reflect the elevated amplitudes of the systolic BCG complexes. A t-test revealed statistically significant differences between amplitude and temporal values (except T RD and A UV ), denoting changes resulting from the sauna bath. Carotid temporal component T RD remained stable, denoting constancy in the ejection pattern of the heart from suppressed respiration. Because of the exercise, a t-test revealed statistically significant differences between amplitude and temporal values (except A HI in the normal part of the respiration of the exercise) denoting changes in temporal and amplitude values due to exercise. Fig. 14. Electromechanical film (EMFi) sensor strips attached to the limbs (dorsalis pedis pulse fromtheankle) andto the carotidarteries onthe neck (Fig. 7; positions 3 and 5). All signals were acquired with the latest version of the mobile physiological signal measurement station (Fig. 5c) using the DAQcard-6036E A/D card. The reason for the decreased BP after the sauna bath is the changed distribution of the blood in the body s veins to lower the body temperature caused by the external heat of the sauna bath. One temperature-decreasing mechanism is vasodilation of the skin blood vessels. Additionally, veins play a major role in body organ temperature control [23]. The number of test subjects was too small to obtain statistically significant results in this study. The PWV value obtained from the left ankle was smaller before the sauna bath accompanying a hypertensive BP value when compared with a higher PWV value after the sauna bath in conjunction with a normal BP value. Usually, when BP decreases, the tension in the arterial walls decreases and causes the pulse wave to travel more slowly (PWV decreases and PTT increases). The PWV was larger after the sauna bath; therefore, the tension of the arterial wall possibly increased in the body after the sauna bath. However, more subjects need to be recorded and analysed to confirm these results. When the frequency domain pictures were studied, the frequency area from 4 Hz to 10 Hz corresponds to the systolic components of the BCG signal. The low-frequency components from the diastolic part of the BCG signal have been merged to the harmonics of the heart rate spectral component (1.1 Hz). In the former BCG studies, the decrease in the J K amplitude of the BCG indicated decreased systemic vascular resistance or aortic wall stiffness [1]. In this study, the J K amplitude decreased after exercise. The reason for the amplitude decrease of the systolic BCG components after exercise (Fig. 10, Table 3) is possibly due to the fact that the recording was conducted 1 h after exercise; the extra heat due to the physical work may have dissipated. Also, diastolic components of the BCG signal have smoothed after exercise (Fig. 10). In the spectral analysis the frequency components after exercise have moved towards the diastolic components of the BCG in time domain (Fig. 11) Conclusions With EMFi sensors, noninvasive measurements in sitting and supine positions are no longer troublesome and time-consuming. Compared with earlier methods, the direct attachment of the sensors on the skin of a subject is easy to implement, thus improving the coupling. In chair measurements, the recoil of the heart is more easily obtained with the flat EMFi sensor located on the chair than with the acceleration sensor. Moreover, the sensing area of the EMFi sensor is greater when compared with the acceleration sensor. Despite that distinction, heart rate determination and PTT measurements were possible with all the sensors examined. The EMFi sensor exhibited considerably higher sensitivity compared with older methods. BCG measurements with acceleration sensors stimulate and encourage the use of this technique. In clinical practice, ECG is the most often used noninvasive method to evaluate patients with suspected ischaemic heart disease. However, an ECG is not sufficiently specific or sensitive to be used in heart disease evaluation without a patient s clinical history [24]. Combination of the ECG and BCG tracings measured with highsensitivity EMFi sensors might add sensitivity without decrease in specificity in suspected ischaemic heart disease evaluation by yielding information about the mechanical function of the heart induced by the electrical stimulus. In conclusion, the constructed device configuration provided reliable measurements of the ECG, BCG, heart sound, carotid, and limb pulse wave signals. The EMFi as a sensor is superior in sensitivity, ease of sensor placement, and comfort on the skin in BCG measurements compared with acceleration sensors. The PWV measurements implemented with the EMFi sensor strips reflected accurately the aortic blood pressure changes. The new EMFi-based BCG measurement system described in this manuscript enables