British Journal of Anaesthesia 84 (1): 38 42 (2000) Behaviour of near-infrared light in the adult human head: implications for clinical near-infrared spectroscopy A. E. R. Young 1 *, T. J. Germon 1, N. J. Barnett 2, A. R. Manara 1 and R. J. Nelson 1 1 Departments of Anaesthesia and Neurosurgery, Frenchay Hospital, Frenchay, Bristol BS16 1LE, UK. 2 Product Development Group, Johnson and Johnson Medical Ltd, Newport, UK *Corresponding author: Department of Anaesthesia, Frenchay Hospital, Frenchay, Bristol BS16 1LE, UK To test theoretical assumptions supporting the use of near-infrared spectroscopy (NIRS) in clinical practice, we examined the behaviour of NIR light transmission and attenuation in the human head. Sterile probes for emitting and detecting NIR light at a fixed separation of 40 mm were placed in turn on intact skin, skull, dura and cerebral cortex of 10 patients undergoing elective neurosurgery. In the first five patients, the detecting probe was moved through successive extracerebral layers with the emitter on the skin surface. In the second five patients, the process was reversed, with the emitting probe moved and the detector in the same place on the scalp. NIR intensity was measured at each tissue interface and compared with the intensity measured at the skin surface with all layers intact. Removal of bone and dura from the light path caused a significant reduction in detected intensity. The largest mean reduction in light intensity was a 14-fold decrease with removal of bone (unadjusted P 0.0001; paired t test). The assumptions that extracerebral tissues contribute little to attenuation of NIR light in the adult head and that most of this attenuation occurs superficially in the scalp are drawn into question by this study. We postulate that the skull and/or its interface with other layers may act as an optical channel, distorting the behaviour of NIR light in the human head. Br J Anaesth 2000; 84: 38 42 Keywords: measurement techniques, near-infrared spectroscopy; brain, near-infrared spectroscopy; surgery, neurological Accepted for publication: July 30, 1999 Reliable non-invasive monitoring of cerebral oxygenation in the concentration of the chromophores absorbing that and metabolism is an important goal in anaesthetic and light. In an organ composed of heterogeneous tissues, intensive care practice. On theoretical grounds, transcutane- absorption of NIR light depends on the proportion of ous reflectance-mode near-infrared spectroscopy (NIRS) the total pathlength which passes through each tissue offers considerable promise in fulfilling these aims. compartment and the change in the concentration of chromophores The principles which underlie the detection of changes in those compartments. in the concentration of oxy- and deoxyhaemoglobin by It has been tempting to assume that there are two principal NIRS are beguilingly simple. Human tissues are partly compartments in the human head: extracerebral tissues and translucent to NIR light. As NIR light passes through tissue, cerebral tissue. Much of the early work in cerebral NIR a proportion of NIR photons are absorbed by chromophores spectroscopy was conducted on the basis that the NIR such as oxy- and deoxyhaemoglobin. The amount of absorp- pathlength in extracerebral tissues was constant and contributed tion depends on two predetermined factors (the wavelength relatively little to NIR light absorption, particularly if of NIR light and the characteristic absorption spectra the emitting source and the detector were widely separated, of oxy- and deoxyhaemoglobin) and two variables (the thus increasing the proportion of pathlength thought to pass concentration of the chromophores in the tissue and the through cerebral tissue. These assumptions may be valid in distance that the photons have to travel from their emitting neonates, but there is increasing evidence that they do not source to the detector the NIR pathlength). The greater apply in adults. 1 4 the pathlength the greater the chance of absorption occurring Schematic representations of NIR spectrometers using and vice versa. single or dual channel sensors often depict an elliptical Thus if the pathlength of NIR light passing through light path between the emitter and detector. 5 However, the a homogeneous tissue remains constant, changes in the presence of discrete layers significantly affects transmission, absorption of NIR light are directly proportional to changes scattering and attenuation of NIR light. 6 9 Little is known The Board of Management and Trustees of the British Journal of Anaesthesia 2000
In vivo behaviour of near-infrared light Fig 1 Apparatus for measuring baseline recording, with both the emitter and detector on the skin surface. Fig 2 Apparatus with the emitter on the skin and the detector on the dura. detector was maintained by the wire coupling at each layer of the transmission of NIR light at tissue interfaces in vivo (Fig. 2). A further 30-s recording of light intensity was such as those between scalp and skull, skull and dura or made at the bone, dura and cerebral cortex. Haemostasis dura and CSF. If, as a result of these interactions, most was meticulous and blood loss negligible during these of the NIR light path in the adult head passes through phases. In the first five studies (group 1), the detecting extracerebral tissues, the capacity of NIRS to detect changes probe was moved progressively through the anatomical in cerebral oxygenation will be reduced, particularly when layers while the emitting probe remained on the scalp. In perfusion and oxygenation of extracerebral tissues are the second five patients, the position of the probes was unstable. reversed (group 2). We have studied, under carefully controlled clinical For each patient, considering light of one wavelength conditions, whether the generally accepted models of NIR within the NIR range (776.5 nm), a mean intensity value light behaviour apply in the human head, and determined for the 30-s recording at each layer was obtained. The the relative effects of the extracerebral tissue layers on the relative light intensity, using the scalp to scalp recording passage of NIR photons. as the baseline measurement, was also calculated for each patient at each extracerebral layer. The ratios were logtransformed Patients and methods and geometric means and 95% confidence intervals (CI) were calculated for groups 1 and 2 and also We studied 10 adult patients (six females), median age for groups 1 and 2 combined. One-sample t tests on the 47 (range 34 65) yr, undergoing elective neurosurgical log-transformed data were used to test the hypothesis that procedures involving a craniotomy or burr hole under there was no change in light intensity at each layer. Paired general anaesthesia. Anaesthesia was induced with propofol, t tests were used to compare the reduction of log-transformed fentanyl and vecuronium and maintained with intermittent light intensity between successive layers. positive pressure ventilation and 1% isoflurane in nitrous oxide and oxygen. Physiological variables, including endtidal carbon dioxide concentration, mean arterial pressure Results and arterial oxygen saturation, were maintained constant Mean recordings of detected light intensity for one wavelength and within normal limits for all patients for the duration of (776.5 nm) at each layer for each patient are shown the study. The study was approved by the Hospital Ethics in Table 1. The geometric mean of light intensity relative Committee and informed consent was obtained from all to a baseline recording (both probes on the skin) are shown patients. for groups 1 and 2 and for all 10 subjects in Table 2. These A sterilizable probe emitting NIR light of four wavelengths data are also shown as averaged values for the three other (776.5, 819.0, 871.0 and 909.0 nm) and a sterilizable wavelengths (819, 871 and 909 nm) for completeness (Table detecting probe (Johnson and Johnson Medical, Cerebral 3). The results for wavelength 776.5 nm are expressed Redox Monitor 2001) were placed on the shaved, cleaned graphically in Figure 3. scalp. The two probes were kept parallel with a stainless Mean light intensity relative to that recorded with both steel wire coupling of 40 mm in length (Fig. 1). NIR probes on the skin (baseline) for all patients was 0.485 intensity was recorded at the detector using a computerized (95% CI 0.223, 1.060) at the bone surface, 0.0338 (0.016, data collection system. An initial baseline measurement 0.071) at the dural surface and 0.0164 (0.005, 0.052) at the with both probes on the scalp was recorded for 30 s. Then, cerebral surface. Although there was an average halving of with one probe remaining in the same position on the scalp, light intensity at the bone surface (with loss of the scalp) the other was placed first onto exposed skull after reflection this was not statistically significant. Significant reductions of the scalp and periosteum, then onto intact dura after were seen with loss of bone and loss of the dura. The craniotomy or burr hole and finally onto the cerebral cortex largest geometric mean reduction was a 14-fold decrease after opening the dura. The separation of emitter and between the bone and dural surfaces with loss of the 39
Young et al. Table 1 NIR intensity for each patient at each layer, averaged over 30 s. Group 1 emitter maintained at the skin and the detector moved through the extracerebral layers, group 2 detector maintained at the skin with the emitter moved through the extracerebral layers Skin to skin Skin to bone Skin to dura Skin to cerebral cortex Group 1 (mean value) Patient No. 1 4018.3 2499.7 76.9 33.3 Patient No. 2 3615.8 793.6 36.5 2.5 Patient No. 3 2527.9 2398.9 271.6 87.1 Patient No. 4 1372.7 698.0 96.3 25.2 Patient No. 5 3445.1 5815.2 722.8 1294.3 Skin to skin Bone to skin Dura to skin Cerebral cortex to skin Group 2 (mean value) Patient A 3419.1 6688.8 225.3 76.2 Patient B 3419.7 234.6 64.6 150.3 Patient C 6312.3 1001.0 145.4 50.6 Patient D 3422.7 3807.7 32.4 22.4 Patient E 3827.8 1063.8 90.9 75.6 Table 2 Geometric mean (95% confidence intervals) values for NIR intensity relative to that obtained with both probes on the skin for groups 1 and 2 and for both groups combined (total), using NIR light of wavelength 776.5nm. 1 One-sample t test, H 0 : geometric mean 1; 2 paired t test Probe position Group 1 Group 2 Total 1 Reduction from previous layer 2 n 5 5 10 Skin and bone 0.6445 (0.2517, 1.6506) 0.3661 (0.0659, 2.0329) 0.4858 (0.2225, 1.0604) P 0.0659 Skin and dura 0.0498 (0.0106, 0.2347) 0.0230 (0.0097, 0.0546) 0.0338 (0.0161, 0.0710) 0.0697 (0.0337, 0.144) P 0.0001 P 0.0001 Skin and cerebral cortex 0.0169 (0.0010, 0.2872) 0.0159 (0.0060, 0.0420) 0.0164 (0.0052, 0.0518) 0.48 (0.2345, 0.9978) P 0.0001 P 0.0494 Table 3 Geometric mean (95% confidence intervals) values for NIR intensity relative to that obtained with both probes on the skin using averaged data from NIR light of wavelengths 819.0, 871.0 and 909.0 nm. 1 One-sample t test, H 0 : geometric mean 1; 2 paired t test skull (unadjusted P 0.0001; paired t test). There was no significant change in measured physiological variables during the study. Probe position Total 1 Reduction from previous layer 2 Discussion Skin and bone 0.6414 (0.2370, 1.7356) In 1977, it was recognized that the differential absorption P 0.3265 of NIR light by haemoglobin, oxyhaemoglobin and Skin and dura 0.0443 (0.0130, 0.1502) 0.0690 (0.0192, 0.2480) cytochrome C oxidase could be used to estimate blood flow P 0.0005 P 0.0017 Skin and cerebral cortex 0.0309 (0.0101, 0.0947) 0.6986 (0.2846, 1.7146) and tissue oxygenation in vivo. 10 Based on these principles, P 0.0002 P 0.03763 cerebral NIR spectroscopy has been used for some years to monitor haemoglobin, oxyhaemoglobin and cytochrome C oxidase in neonates whose heads transmit sufficient NIR light. 11 In older children and adults, transmission NIRS is prevented by the thickness of the scalp and skull, and myelination of the brain. Reflectance-mode NIRS was developed to overcome these problems. Using this technique, NIR light is collected from the surface of the scalp ipsilateral to the light source so that the photon pathlength is significantly reduced. NIR light is thought to follow an arc whose depth is proportional to the distance between the emitter and detector and will, in theory, travel through scalp, skull, dura and finally brain before returning through the same tissue layers to be detected at the skin surface. 5 The path of NIR light through the human head and thus the volume of cerebral tissue interrogated by reflectance- Fig 3 Mean fractional change in detected light intensity relative to baseline (with both probes on the skin) after loss of successive layers of the human head. mode NIRS is unknown. The pathlength of the emitted NIR light must be known if absorption spectra are to be used to 40
In vivo behaviour of near-infrared light optical pathlength were estimated in adult subjects with and without a scalp tourniquet. 20 No change was observed in NIR pathlength or CBV after arrest of scalp blood flow, confirmed by laser Doppler velocimetry. The authors con- cluded that scalp blood flow changes have little or no effect on NIRS measurement of cerebral haemodynamics. Cerebral blood flow (CBF) in anaesthetized patients has also been calculated using NIR probes placed first on the scalp and then on the dura. 4 CBF measured through the intact scalp was less than that measured using xenon 133 and was one- third that of NIRS measurement of CBF with probes placed on the dura. The authors concluded that underestimation was a result of extracerebral attenuation of NIR light. They suggested that 60 70% of the total optical pathlength of NIR light in the human head was through the extracerebral tissues, but they did not comment on the relative contribu- tions of the scalp and skull or on whether the dura or CSF might have an effect. Another recent in vivo NIR study demonstrated that an increase in interoptode distance from 0.7 to 5.5 cm increased the sensitivity of the monitor to cerebral attenuation changes. However, at any interoptode distance, extracerebral changes were significant compared with cerebral changes, implying that extracerebral effects are important in the use of clinical NIR spectroscopy monitors, particularly at shorter interoptode distances. 21 We studied transmission of NIR light in two distinct ways. In the first, a detecting probe was moved sequentially through the extracerebral layers of the cranium so that transmitted light intensity was sampled at each tissue interface. If the traditional model of NIR light transmission is correct, that is light travels in an arc whose depth is dependent on the separation of the emitting and detecting probes and that the extracerebral tissue layers have little effect on NIR light attenuation and path length, we would have anticipated small increases in detected optical intensity with removal of each successive extracerebral layer from the NIR light path. Alternatively, in the second part of the study, as the emitter was moved through the extracerebral layers, a greater proportion of the NIR pathlength would pass through cerebral tissue with an increased attenuation and we would have anticipated successive reductions in optical intensity detected at the skin surface. In contrast with the anticipated findings, we observed a major decrease in detected optical intensity when bone was removed from the NIR path in both arms of the study. The changes observed when the scalp was removed from either the emitting or detecting pathway varied considerably from patient to patient and the mean reduction in intensity was not statistically significant. There was a small, statistically significant decrease in detected light intensity when the dura and CSF were removed. We postulate that if all the extracerebral layers are intact, there is a preferential path or channel for NIR light from emitter to detector. That is, the bone or bone dura interfaces, and to a lesser extent the dura and CSF, may be allowing calculate absolute quantities of physiological chromophores. The mean photon pathlength may be estimated using timeof-flight and other methods. 12 15 These use sophisticated technology which is not yet suitable for bedside monitoring. The effect of the extracerebral layers on the passage and pathlength of NIR light is poorly understood. There is evidence that extracerebral attenuation of NIR light influences the accuracy of all cerebral spectrometers. 1 4 It is likely that the biological and optical characteristics of the extracerebral layers vary from patient to patient and from time to time. Indeed, it is theoretically possible for NIR light to travel from the emitter to the detector without passing through cerebral tissue. It has still to be shown that the configuration of a single NIR source and one or more ipsilateral receivers is capable of detecting intracranial NIR light attenuation changes through the intact scalp, skull, dura and CSF of adult humans. In the development of clinical NIRS monitors, extracerebral attenuation of NIR light has been assumed to be small and/or predictable. In particular, bone has been thought to be translucent to NIR light. Despite concerns about the effect of extracerebral tissues on NIR light, there have been few in vivo studies and most published work concerns the use of mathematical and in vitro models to predict the behaviour of NIR light in the adult human head. 12816 18 In these phantom models, the amount of NIR light penetrating cerebral tissue is dependent on the separation of the light emitter and detector and on the optical characteristics of the extracerebral layers. The effect of discrete anatomical layers has been modelled. Using a simple Monte Carlo simulation of the adult head comprising two concentric spheres of differing media, Hiraoka and colleagues calculated that the pathlength of NIR light in cerebral tissue was 40 55% of the total pathlength 17. Others have studied the effect of the CSF layer in a simple model with layers of different optical properties and in a more complex model simulating cerebral sulci. 19 Mean optical pathlength was calculated experimentally and compared with Monte Carlo predictions. The results of that study suggested that at an emitter detector separation of 5.0 cm, 55% of the NIR light pathlength was in the scalp and skull, 20% in the CSF and, in contrast with Hiraoka s work, only 15% of the NIR pathlength was in the cerebral cortex. Because CSF has low scattering and absorption coefficients, the authors postulated that the CSF is acting as a conduit or channel for the light. In their study, however, the scalp and skull were considered as one homogeneous layer. In practice, the scalp and skull have very different optical properties and thus have very different effects on NIR light behaviour and path shape. The authors stated that the fractional pathlength of scalp and skull is much greater (55%) than that of any other layer. This would support our findings; since the absorption coefficient of the skin is much greater than that of the skull, most light should therefore pass unattenuated through the skull. In another study, cerebral blood volume (CBV) and 41
Young et al. optical channelling of NIR photons from emitter to 4 Owen-Reece H, Elwell CE, Harkness W, et al. Use of neardetector. Mathematical models have suggested that the CSF infrared spectroscopy to estimate cerebral blood flow in conscious and anaesthetised adult subjects. Br J Anaesth 1996; 76: 43 8 layer may act as an optical channel. 19 We do not know if 5 Cope M, Delpy DT, Reynolds EOR, Wray S, Wyatt JS, Van der the decrease in light intensity that we observed was a result Zee P. Methods of quantitating cerebral near infrared spectroscopy of loss of the bone layer itself or loss of the bone dura data. Adv Exp Med Biol 1988; 222: 183 9 interface. 6 Firbank M, Schweiger M, Delpy DT. Investigation of light piping We accept that it was impossible to avoid some distortion through clear regions of scattering objects. Proc Soc Photooptical of boundary conditions after craniotomy or burr hole. Instrumentation Eng 1995; 2389: 167 73 We also accept that there are minor increases in the 7 Firbank M, Arridge SR, Schweiger M, Delpy DT. An investigation of light transport through scattering bodies with non-scattering emitter detector distance (of the order of a few millimetres) regions. Phys Med Biol 1996; 41: 767 83 with movement of the probes from the scalp to the cerebral 8 Hielscher AH, Liu H, Wang L, Tittel FK, Chance B, Jacques SL. surface. However, the consistency and magnitude of the Time resolved reflectance measurements on layered tissues with results in both arms of the study suggest that optical strongly varying optical properties. Proc Soc Photooptical channelling may be a real phenomenon and not an artefact Instrumentation Eng 1995; 2323: 351 9 arising from disturbance of light transmission at optical 9 Okada E, Firbank M, Schweiger M, Arridge SR, Cope M, Delpy DT. A theoretical and experimental investigation of the effect of boundaries. It is difficult to envisage another clinical experisulci on light propagation in brain tissue. Proc Soc Photooptical mental technique in which it is possible to maintain boundary Instrumentation Eng 1995; 2626: 2 8 conditions. Furthermore, if the performance of NIRS is so 10 Jobsis FF. Non-invasive infrared monitoring of cerebral and sensitive to boundary conditions, it will be unreliable myocardial oxygen sufficiency and circulatory parameters. Science in those conditions such as head injury, subarachnoid 1977; 198: 1264 6 haemorrhage and postoperative monitoring when the boundcerebral blood and tissue oxygenation on newborn infants by 11 Cope M, Delpy DT. System for long-term measurement of ary conditions are altered by the clinical condition itself. infrared transillumination. Med Biol Eng Comput 1988; 26: 289 94 We have only considered three major extracerebral layers. 12 Delpy DT, Cope M, van der Zee P, Arridge S, Wray S, Wyatt J. We have not been able to study the CSF layer independently Estimation of optical pathlength through tissue from direct time of the other layers and it is this layer which is most likely of flight measurement. Phys Med Biol 1988; 33: 1433 42 to change with small alterations in intracranial pressure not 13 Lakowicz JR, Berndt K. Frequency-domain measurements of accompanied by major changes in cerebral oxygenation. photon migration in tissues. Chem Phys Lett 1990; 166: 246 52 14 Cope M, Delpy DT, Wray S, Wyatt JS, Reynolds EOR. A CCD Even under carefully controlled conditions, it is clear that spectrometer to quantitate the concentration of chromophores extracerebral tissues have a profound influence on NIR in living tissue utilising the absorption peak of water at 975nm. light transmission, and the behaviour of NIR light in these Adv Exp Med Biol 1989; 248: 33 40 anatomical layers must be understood fully if NIRS is to 15 Matcher SJ, Cope M, Delpy DT. Use of the water absorption be developed as a reliable clinical monitor. spectrum to quantify tissue chromophore concentration changes in near infrared spectroscopy. Phys Med Biol 1994; 39: 177 96 16 Hielscher AH, Liu H, Wang L, Tittel FK, Chance B, Jacques SL. Acknowledgements Determination of blood oxygenation in the brain by time-resolved We thank Dr John Carlin (Clinical Epidemiology and Biostatistics Unit, reflectance spectroscopy: influence of the skin, skull and meninges. Royal Children s Hospital, Melbourne, Australia) for statistical advice and Proc Soc Photooptical Instrumentation Eng 1994; 2136: 1 27 Dr Paul Brown (Department of Medical Physics, Frenchay Hospital) for 17 Hiraoka M, Firbank M, Essenpreis M, et al. A Monte Carlo help with data collection. Dr A. E. R. Young was supported by a grant investigation of optical pathlength in homogenous tissue and its from Research and Development Directorate, NHS Executive, South and application to near-infrared spectroscopy. Phys Med Biol 1993; 38: West Region. 1859 76 18 Okada E, Firbank M, Delpy DT. The effect of overlying tissue on the spatial sensitivity profile of near-infrared spectroscopy. Phys References Med Biol 1995; 40: 2093 108 1 Germon TJ, Young AER, Manara AR, Nelson RJ. Extracerebral 19 Cui W, Kumar C, Chance B. Experimental study of migration absorption of near-infrared light influences the detection of depth for the photons measured at sample surface. Proc Soc increased cerebral oxygenation monitored by near-infrared Photooptical Instrumentation Eng 1991; 1431: 180 91 spectroscopy. J Neurol Neurosurg Psychiatry 1995; 58: 477 9 20 Owen-Reece H, Elwell CE, Wyatt JS, Delpy DT. The effect of 2 Harris DNF, Bailey SM. Near infrared spectroscopy in adults. scalp ischaemia on measurement of cerebral blood volume by Anaesthesia 1993; 48: 694 6 near-infrared spectroscopy. Physiol Meas 1996; 17: 279 86 3 Hopton P, Walsh TS, Lee A. CBF in adults using near-infrared 21 Germon TJ, Evans PD, Barnett NJ, Wall P, Manara AR, Nelson RJ. spectroscopy: potential for bedside measurement. Br J Anaesth 1996; 77: 131 Cerebral near-infrared spectroscopy: emitter detector separation must be increased. Br J Anaesth 1999; 82: 831 7 42