Main

Disturbances of cerebral blood volume (CBV) in the immature brain have recently been proposed to be predictive of subsequent brain injury in infants suffering birth asphyxia(1), thereby aiding the development of clinical interventions directed at protective therapy. Levels of cerebral blood flow, by contrast, appear to have poorer prognostic value(1). Thus, CBV values twice the normal value were measured in brain-injured infants(2), whereas, in contrast, cerebral blood flow values lay within the normal range(1). Coupled with elevated CBV, the capacity to increment CBV on elevating arterial carbon dioxide tension (PaCO2), known as CBV responsiveness, is reduced or lost in brain-injured infants(3), a phenomenon that has been termed "vasoparalysis"(4).

Recently the diagnostic value of CBV measurement has been assessed using near-infrared spectroscopy (NIRS), a technique for measurement of blood volume in the brain that is particularly attractive for use in the neonatal intensive care unit because it is noninvasive, nonradioactive, and portable. NIRS has been used in a number of studies of neonatal cerebral hemodynamics(57). However, validation of NIRS is required before its promise can be fully realized inasmuch as there are several physical and physiologic assumptions required for NIRS measurements in the brain that may affect its accuracy(8,9). Moreover, a recent comparison of CBV obtained using NIRS against measurements made with 99mTc-labeled erythrocytes in a newborn animal model failed to demonstrate a correlation between the methods(10), despite earlier validation of NIRS made using plethysmography(11). There is also uncertainty about normal values of CBV and its regulation in the immature brain, because CBV values measured with NIRS in normal infants appear to be much less than those in adults obtained with different methods(12,13).

Because these methodological uncertainties limit the potential usefulness of CBV and CBV responsiveness in the care of critically ill newborn infants, we aimed in this study to compare CBV values measured by NIRS with those obtained using radiolabeled red blood cells (RBC) and radiodinated serum albumin (RISA), methods that are considered to be the "gold standard" for blood volume estimation(10,14,15). In addition, because no validations of NIRS exist for the immature brain, we aimed to establish normal values for total CBV using an immature lamb at midgestation when the brain is similar to that of a 30-wk gestation human infant(16). Finally, because there is substantial variability in cerebral blood flow between regions of the immature brain with differing susceptibilities to asphyxial brain injury(17,18), and because the NIRS estimate is dependent on the presence of Hb, we also examined regional variations in RBC, plasma, and total CBV using radiolabels.

METHODS

Animal preparation. Studies were performed on a total of 18 fetal lambs of Merino-Border Leicester cross ewes with time-dated pregnancies (92 ± 1 d, mean ± SEM; term = 147 d). Animals were used in three studies, each using a separate group of animals: a study using radiolabels to measure global and regional CBV (n = 10); a second study using NIRS to measure global CBV in a separate group (n = 5); and an additional, preliminary study to determine the mixing characteristics of injected radiolabels in the fetal circulation in a third group (n = 3). All procedures were performed in accordance with the guidelines established by the National Health and Medical Research Council of Australia, and were approved by the Ethics in Animal Experimentation Committee of Monash University. Under general anesthesia (25 mg/kg α-chloralose and 5 mg/kg ketamine hydrochloride for induction and 15-25 mg ·kg-1·h-1 α-chloralose for maintenance) the pregnant ewe was intubated and ventilated (40% O2, balance N2). The uterus was exposed through a midline abdominal incision, and an incision was made to allow delivery of the head, upper limbs, and a hindlimb of the fetus. Polyvinyl catheters (2 mm OD) filled with a saline-heparin solution (50 IU/mL) were inserted bilaterally into the axillary artery to enable monitoring of fetal arterial pressure and extraction of blood samples. Additional catheters for isotope infusion were inserted nonocclusively into a jugular vein and a lateral superficial hindlimb vein and advanced so that their tips were in the superior and inferior venae cavae, respectively. Finally, a catheter was attached to the fetal thorax at atrial level for amniotic fluid pressure measurement. The fetus was then partially returned to the uterine cavity, leaving the head and a forelimb outside the uterus for NIRS and pulse oximeter access, respectively. The head and forelimb were covered with a saline-soaked, warmed towel to protect the skin from evaporation, and the fetus was positioned to ensure that the umbilical circulation was not compromised.

Physiologic measurements. Fetal axillary artery pressure and amniotic fluid pressure were measured with calibrated strain-gauge pressure transducers (model 120B, Hewlett-Packard, Waltham, MA) connected to signal conditioners (model 800Z, Neomedix Systems, Sydney, New South Wales, Australia). Arterial pressure was averaged electronically and referenced to amniotic fluid pressure. Heart rate was derived from a tachometer (model NT122, Neomedix Systems) triggered by the arterial pressure pulse. These signals were monitored throughout the study and displayed on an eight-channel recorder (model 800Z, Neomedix Systems). Arterial blood pH, arterial oxygen tension (PaO2), and PaCO2 were measured in 200-µL samples collected in heparinized plastic syringes and analyzed immediately (model ABL500, Radiometer, Copenhagen, Denmark). Arterial blood Hb concentration and Hb oxygen saturation (SaO2) were measured in duplicate in 20-µL samples using a hemoximeter (model OSM2, Radiometer).

CBV measurements using NIRS. Optodes (NIRO-500 Hamamatsu Photonics K.K., Hamamatsu City, Japan) were secured bilaterally on either side of the fetal skull over the temporal region and covered with a lightproof dressing. The mean interoptode distance was 3.9 ± 0.2 cm. A differential path-length factor of 4.55 was used for calculations(19). Beat-to-beat fetal SaO2 was measured at the fetal forelimb with a pulse oximeter (Nellcor 180, Nellcor Incorporated, Pleasanton, CA). Total CBV (mL/100 g) was determined by measuring the effect on cerebral oxyhemoglobin and deoxyhemoglobin concentrations as fetal SaO2 was lowered by 11 ± 2% from the baseline level during 1-3 min(2). The reduction in fetal SaO2 was achieved by reducing the ewe's inspired oxygen concentration to 8-12% without alteration in end-tidal CO2. Fetal pH, PCO2, heart rate, and mean arterial pressure were unchanged during the measurements. CBV (mL/100 g) was calculated from data averaged during 10-s periods using commercially available software (Onmain, Hamamatsu Photonics K.K.) that incorporates the following formula(2): CBV = (Δ[HbO2] - Δ[Hb])/(2 × ΔSaO2 × H × R) where Δ[HbO2] is the change in cerebral oxyhemoglobin, Δ[Hb] is the change in cerebral deoxyhemoglobin, ΔSaO2 is the change in oxygen saturation, H is Hb concentration, and R is the cerebral-to-large vessel hematocrit ratio. The coefficient of variation of the NIRS measurement calculated from 12 repetitions in three animals was 6%.

CBV measurements using radiolabels. We used procedures similar to those that have been described previously for preparation of radiolabeled RBC and RISA(14). Fetal blood (2.5 mL) was withdrawn into 1 mL of acid-citrate-dextrose solution and labeled with chromium 51 by incubating the blood with Na251CrO4- (New Research Products, DuPont, MA) for 30 min at 40°C. The withdrawn fetal blood was replaced with an equal volume of maternal blood. After incubation, the specimen was gently centrifuged at 3000 rpm for 5 min, and the supernatant was discarded and replaced with an equal volume of warmed normal saline. This centrifugation-washing procedure was performed three times to remove free chromium 51. Ten microcuries (0.2 mL) of RISA (125I-labeled albumin, Amersham International, Bucks, UK) was added to the RBC mixture for use as a plasma volume marker. Less than 5% of the iodine 125 existed in the free form in the RISA preparation. After counting radioactivity of 51Cr-RBC and 125I-RISA in the mixture, it was kept in a water bath at 40°C until injection.

RBC and plasma volume determinations. The 51Cr-RBC-125I-RISA mixture was injected into the fetus during 10-15 s. To accelerate mixing within the fetal circulation, two-thirds of the mixture was injected into the inferior vena cava, while simultaneously one-third was injected into the superior vena cava(14). Blood volume reference samples of approximately 0.3 mL were collected from the axillary artery catheter at 1, 2, 3, and 5 min after the injection (n = 10). After measurement of their hematocrit, the blood samples were dispensed into preweighed vials and weighed, and their radioactivity was counted. Throughout the blood collection procedures maternal blood was infused i.v. to the fetus to exactly replace the withdrawn blood. Five minutes after injection of the blood volume tracers, a time sufficient to allow complete mixing (Fig. 1), the fetus (body weight, 733 ± 29 g; n = 10) was killed with an overdose of sodium pentobarbitone (100 mg/kg), a ligature was firmly tightened around the neck, and the animal was quickly decapitated.

Figure 1
figure 1

Volumes of distribution (mean ± SEM) for 51Cr-RBC and 125I-RISA in fetal lambs (n = 3) after i.v. injection at zero time. Note that the period 0-5 min corresponds to mixing of the indicators within the circulation, followed by 1) a plateau in the curve for 51Cr-RBC significantly completeness of mixing and stability of radiolabeled RBC within the circulation after 5 min; and 2) conversion of the 125I-RISA curve to a steady increase explained by loss of indicator from the circulation that continues after mixing is complete.

Brain tissue sample analysis. Immediately after the fetus was decapitated, the head was immersed in isopentane cooled to -140°C with liquid nitrogen. When freezing was complete (5 min), the brain was removed from the skull while still frozen, wrapped in plastic to prevent desiccation, and stored at -40°C. Subsequently, the lamb brain was weighed (14.9 ± 0.5 g, n = 10) and, while still frozen, dissected into the following regions: cortex, midbrain, brain stem, cerebellum, white matter, and choroid plexus. The sagittal and transverse sinuses were removed during the dissection, but all other blood vessels on the brain surface were included. The dissected samples, along with the tissue remains, were placed in preweighed plastic vials, weighed, and counted for radioactivity.

Radioactivity measurements. The 51Cr-RBC-125I-RISA mixture, blood reference samples, and regional brain tissue samples were counted in an auto-gamma scintillation spectrometer with automated computer correction for spectral overlaps (model 1282 Compugamma, LKB-Wallac, Turku, Finland). Samples were counted for a time sufficient to limit the counting errors to less than 10%. In the white matter samples, the levels of 51Cr-RBC were too low to achieve this accuracy despite prolonged counting times; in these samples RBC volume was deemed to be undetectable and assigned a zero value.

Calculations. Two methods of calculating systemic blood volume (SBV, mL) of the fetus were used. In the first (single-label method) we used the general relation SBV = Cinj/Cref where Cinj is the injected radioactivity (51Cr-RBC or 125I-RISA, counts/min) and Cref is the radioactivity of the reference blood sample (counts·min-1·mL-1). The single-label method was applied to each of the radiolabels, thereby providing two SBV estimates. In preliminary studies extending for 60 min, we established that indicator mixing was complete within 5 min (Fig. 1). Subsequently there was a steady loss of the 125I-RISA plasma label from the circulation that was corrected by extrapolating the 5- to 60-min values back to zero time to obtain the plasma distribution volume (Fig. 1). Although no correction was required for 51Cr-RBC as their volume of distribution was steady after 5 min of mixing, these values were also extrapolated back to zero time. Volumes of the reference blood samples were calculated as weight (g) divided by blood density (g/mL). Blood density was estimated as 1.01 + 0.07 × hematocrit of the reference sample, assuming values of 1.01 g/mL and 1.08 g/mL, respectively, for the density of plasma and RBC(14). A third estimate of SBV was also calculated by summing separate estimates of systemic RBC and plasma volumes (double-label method). RBC volume was calculated by multiplying the value of SBV derived from 51Cr-RBC by the hematocrit of the reference sample; systemic plasma volume was calculated by multiplying the value of SBV derived from 125I-RISA by (1 - hematocrit).

Blood volumes in cerebral tissue samples (CBV, mL/100 g) were calculated from the relation CBV = Ct/Cref where Ct is the radioactivity (51Cr-RBC or 125I-RISA) per 100 g tissue (counts·min-1·100 g-1) and Cref is the radioactivity of the 5-min reference blood sample (counts·min-1·mL-1). In a double-label method, RBC and plasma volume were estimated separately, and from these estimates, whole blood volumes, together with hematocrit values, were obtained for each brain region; values for the whole brain were obtained by summing the values of all brain samples. Two additional estimates of total brain blood volume were made using the total brain radioactivity and the 5-min reference sample radioactivity of each of the radiolabels (single-label method); these calculations used hematocrit values for brain tissue that we calculated using the measured arterial hematocrit and the average value for the cerebral-to-large vessel hematocrit ratio that we had previously determined using the double-label method.

Statistics. Nonparametric tests were used for statistical comparisons (Sigmastat, Jandel Corporation, San Rafael, CA). The Mann-Whitney rank sum test was used to test for differences between the baseline physiologic characteristics of the NIRS and radiolabel groups and between the estimates of CBV made with the two methods. The Wilcoxon signed rank test was used to compare the paired estimates of SBV. Regional values of CBV and hematocrit were compared with an ANOVA for repeated measures on ranks (Friedman ANOVA) followed by a Student-Newman-Keuls test to identify the specific regions that differed. A probability of p ≤ 0.05 was considered significant. Data are presented as mean ± SEM.

RESULTS

Physical characteristics and cardiorespiratory status of fetal lambs used in the NIRS studies were essentially identical to those of lambs used for the radiolabel studies (Table 1). In both groups, heart rate, arterial pressure, Hb, and blood gas and pH status are within normal ranges for fetal lamb preparations of 90-100 d gestation(17,20,21).

Table 1 Baseline characteristics of immature fetal lambs in the two measurements groups

Values of SBV (Table 2) determined with double or single radiolabel methods were not different. SBV averaged 108 ± 15 mL/kg for the single-label 51Cr-RBC method, 124 ± 8 mL/kg for the single-label 125I RISA method, and 118 ± 11 mL/kg when the RBC and plasma volumes determined separately by the double-label method (51Cr-RBC and 125I-RISA) were added to obtain the whole blood volume.

Table 2 Systemic blood volumes (RBC, plasma, and total blood) in immature fetal lambs (n = 3)

Total CBV determined by NIRS averaged 2.5 ± 0.2 mL/100 g. This was essentially identical to the values obtained using radiolabels (Table 3). These averaged 2.5 ± 0.2 mL/100 g for the double-label value obtained by adding the RBC and plasma volumes that were obtained separately, 2.5 ± 0.4 mL/100 g for the single-label 51Cr-RBC method, and 2.5 ± 0.2 mL/100 g for the single-label 125I-RISA method.

Table 3 Blood volumes of the whole brain measured in immature fetal lambs using radiolabels (n = 10) or NIRS (n = 5)

Regional tissue levels of plasma, RBC, and whole blood (mL/100 g) varied significantly within the brain (Table 4). Whole blood volume was largest in choroid plexus (16.2 ± 2.1 mL/100 g) and least in white matter (0.7 ± 0.1 mL/100 g). A significant hierarchy (p ≤ 0.05) existed among regions with the order from highest to lowest values of whole blood volume being choroid plexus > cerebellum > cortex > midbrain = brain stem > white matter. There was a similar hierarchy (p < 0.05) of regional plasma volumes, which were highest in choroid plexus (13.0 ± 1.6 mL/100 g), and least in white matter (0.8 ± 0.1 mL/100 g). A hierarchy of regional RBC volume was also present, with the value in choroid plexus (3.2 ± 0.9 mL/100 g) significantly exceeding the values for the midbrain and brain stem, which in turn exceeded that of the white matter, in which RBC were undetectable. Reflecting the RBC distribution, the regional brain tissue hematocrit was least in white matter (Table 4). Regional hematocrit differed from the hierarchy of blood volume distributions in other respects, being more uniform overall, and in having the highest value in the cortex (0.27 ± 0.03 mL/100 g).

Table 4 Regional volumes of whole blood, plasma, and RBC in the immature fetal lamb brain

DISCUSSION

Our study is the first to assess the use of NIRS for quantifying blood volume within an immature brain. Our results clearly validate the use of NIRS for measuring CBV, inasmuch as this technique provides values that are essentially identical to those obtained using radiolabeled tracers (51Cr-RBC and 125I-RISA). Using NIRS we measured a CBV of 2.5 mL/100 g of brain, a value equal to that we obtained using the two radiolabeled tracers. Furthermore, the measurements are similar to the value of 2.2 ± 0.4 mL/100 g obtained using NIRS in newborn infants(2). These values are substantially less than the mean CBV of 4.8 ± 0.4 mL/100 g obtained in adult humans using single-photon emission tomography(12), a difference that may be age-related. However, the lower CBV in the immature brain most likely reflects a lower vascular density throughout the brain(17,18) as well as a larger proportion of white matter in which blood volumes are very low (Table 3).

Our validation of the NIRS method for measurement of CBV contrasts with the failure to demonstrate a correlation between NIRS and radiolabel-based CBV estimates in the only other published study comparing the methods. In that study, Brun et al.(10) recorded an absolute cerebral erythrocyte volume in newborn piglets that amounted to approximately one half that obtained with 99mTc-labeled erythrocytes. In addition, the cerebral hematocrit measured by these authors using NIRS exceeded the large vessel hematocrit, a finding seemingly inconsistent with the low Hb value found with NIRS in their study, and also at variance with published data showing that the hematocrit is lower in cerebral tissue than in large vessels because of the Fahraeus effect(15). Beyond the obvious differences in the animal model and the manufacturer of the NIRS machines used for the studies we are uncertain as to explanations for the discrepancies between our study and that of Brun et al.(10).

Because SaO2 was lowered for the NIRS measurement, but not for the radiolabel measurement, the two groups are different in terms of oxygenation at the point of measurement, and the concern exists that lowering arterial oxygen levels as part of the NIRS method might alter the level of CBV that is being estimated(2). However, the identical values of CBV obtained with the two methods in our study supports the argument that manipulation of oxygen levels for CBV determination with NIRS causes minimal changes in the immature cerebral circulation(2).

Our experiments revealed significant regional differences in CBV, with the choroid plexus having the largest, and the white matter having the smallest, blood volume in a distinctive regional hierarchy of values that was repeated for RBC, plasma, and whole blood. This distribution of the blood volumes is similar to the regional distribution of cerebral blood flow in the fetal lamb(17), suggesting that regional differences in vascularity largely underpin the physiologic differences in the immature cerebral circulation. Notably, extremely low values of RBC volume parallel the low blood flow levels of white matter(17,21). Low RBC volumes are reflected in low values of hematocrit (Table 4), a characteristic of white matter structures that is also evident in the mature brain(15). Because the NIRS measurements of CBV are based on the detection of Hb, our data suggest that among the brain regions most susceptible to hypoxic-ischemic insult, white matter is unlikely to make a significant contribution to the whole-brain CBV estimated with the technique.

We used as an experimental model the midgestation lamb fetus, an animal that has found widespread application in physiologic investigations of the immature cerebral circulation(17,1921) although, until now, not for study of blood volume regulation. Importantly, the midgestation lamb has a cerebrovascular anatomy that resembles that of the preterm human brain at 30 wk of gestation, particularly in that it has an extensive germinal matrix(16), making it a valuable model of intracranial hemorrhage(16) and hypoxic-ischemic brain damage(22). The animals in this study were in a normal, stable cardiorespiratory state as evidenced by values of blood gases, heart rate, and blood pressure that are within the normal range for the lamb fetus at this developmental stage(17,20,21). Accordingly, we consider the unique values of CBV and SBV obtained in this study to represent normal baseline values for the lamb at this developmental stage.

We measured RBC and plasma volumes using radiolabel techniques that have been carefully evaluated for measuring SBV in the fetal lamb(14). The methodological considerations that are considered critical to the accuracy of this technique for measuring blood volume are 1) that the amount of unbound label in the injected mixture is negligible; 2) that uniform mixing of the radiolabeled indicators is promptly established within the circulation after their injection; and 3) that there is appropriate correction for the loss of labels from the circulation. We ensured that no free chromium 51 existed in the RBC mixture by repeated washing during the labeling procedure. Furthermore, fetal RBC were used for the labeling process because, unlike maternal RBC, autologous fetal cells labeled with chromium 51 are not removed from the fetal circulation and so do not erroneously increase the chromium 51 distribution volume(14).

To facilitate mixing of the indicators after venous injection, we used the method of Brace(14) in which the injectate is divided between the inferior and superior venae cavae in volume proportions equal to the relative blood flow rates in the vessels. In our study complete mixing of radiolabels in the circulation was accomplished after 5 min of recirculation, as shown by the form of the blood volume-time relationships (Fig. 1). Within the first 5 min after injection, there was a steady increase in the volumes of distribution for both plasma and RBC that we interpret as reflecting the progressive distribution of labels within the circulation. Subsequently the relationship for RBC exhibits a plateau between 5 and 60 min, signifying completion of the mixing process. Importantly, the plateau in the curve also shows that labeled RBC are not lost from the circulation during the experimental period, so that estimates of both systemic and cerebral tissue RBC volumes are not time-sensitive after the achievement of complete mixing. However, in the case of the plasma label, there was evidence of a continuing loss of 125I-RISA from the circulation after the mixing phase was complete because its volume of distribution maintained a steady rise between 5 and 60 min (Fig. 1). SBV estimates were corrected for this loss of the plasma label by the usual procedure of zero-time extrapolation(14), but this correction was not feasible in the case of tissue plasma volume estimates because tissue radioactivity measurements were made at a single time corresponding to the death of the animal. Accordingly, in our experiments we arrested the circulation and decapitated the animal 5 min after injection of the radiolabels, a time that was optimal in that it ensured complete mixing of both labels while minimizing loss of the plasma label from the circulation.

SBV in these immature lambs averaged 108 ± 15 mL/kg when 51Cr-RBC was used as the indicator and 124 ± 8 mL/kg with 125I-RISA. Using the double-indicator method (51Cr-RBC and 125I-RISA), SBV was 118 ± 11 mL/kg in these immature lamb fetuses. Published SBV values in late-gestation fetuses are in close agreement when 51Cr-RBC is used, averaging approximately 110 mL/kg fetal body weight(14,22), a value close to the one we report for the immature fetus. Published values obtained using 125I-RISA, however, are widely variable and exceed the 51Cr-RBC value(14,23). As reviewed in detail by Brace(14), higher estimates are largely explained by failure to correct for loss of the indicator from the circulation using zero-time extrapolation. The close concordance of the two methods in our study suggests that little error was introduced by immediate losses of free iodine 125 into the interstitium, and that the method of zero-time extrapolation adequately corrected for the slow loss of 125I-RISA from the circulation.

CBV, as well as its regional distribution, is influenced by the method of tissue collection and processing(15,24). Rapid freezing of the brain is preferred over decapitation without freezing because it prevents blood loss(15). Additionally, in situ fixation is superior to removal of the brain from the skull because it preserves the regional distribution of blood(24). Because NIRS measurements include blood within the entire cranial vault, we aimed to prevent blood loss from the brain before measurement of the radiolabel contents. We were also concerned that the regional distribution of CBV should be preserved during the dissection and processing of cerebral tissue. Accordingly, a ligature was placed firmly around the neck before decapitation and freezing of the brain in situ, and the brain was removed from the skull and dissected into regions for radioactivity measurements while still frozen.

In terms of the accuracy of the measurement of CBV, the absence of any loss of 51Cr-RBC from the circulation after the 5-min mixing period (Fig. 1) makes it the indicator of choice for estimating cerebral tissue blood volumes. Use of 125I-RISA may be less accurate if significant accumulation of tracer were to occur in brain tissue. We observed an increase in the 125I-RISA distribution volume of approximately 30% in the hour after the initial mixing phase, a rate that translates into approximately 3% loss of indicator during the 5-min circulation time before decapitation and brain freezing. Because of the blood-brain barrier, loss into the brain parenchyma can reasonably be expected to be less than into other tissues. Single- and double-label methods represent alternative methods of measuring total CBV. The single-label method offers the advantage of simplicity because just one label (either 51Cr-RBC or 125I-RISA) is required. However, this method suffers the disadvantage that a value for cerebral hematocrit, which is well-recognized to be significantly lower than large vessel hematocrit(15), must be assumed for calculation of total CBV from the RBC or plasma volume. The strength of the double-label method is that no assumption of cerebral hematocrit is needed because this is measured. The method also allows determination of whole brain as well as regional cerebral hematocrit. In our study, total CBV obtained by summing the individual estimates of RBC and plasma volume averaged 2.5 mL/100 g of brain, similar to the values obtained from the single-indicator estimates that we calculated using a cerebral hematocrit value that was determined using the double label. The close concordance of these values suggests that use of 125I-RISA, along with the assumption of a cerebral-to-large vessel hematocrit ratio of 0.8, is also a valid indicator for measurement of CBV in the immature lamb brain.

In summary, our experiments in the immature lamb brain have demonstrated a close concordance between NIRS-derived CBV measurements and values obtained with traditional gold-standard radiolabel techniques. Thus CBV is accurately measured by NIRS in this animal model. Further studies of NIRS in other aspects of cerebrovascular physiology and pathophysiology, such as the changes occurring during hypoxic-ischemic injury, are needed to assess the accuracy and clinical usefulness of this technique for the neonatal intensive care unit.