Improvements in neonatal intensive care have reduced the mortality rate associated with preterm birth, but unfortunately these infants remain at a high risk of neurological complications, including learning disabilities and cerebral palsy.1 Preterm infants are vulnerable to ischemic and hemorrhagic brain injury in part because of an underdeveloped cerebral vasculature, including limited or impaired autoregulation.2 As a result, unstable arterial blood pressure could lead to dangerous fluctuations in cerebral blood flow (CBF). Recent studies using near-infrared spectroscopy (NIRS) to monitor cerebral blood oxygenation have reported that periods of impaired autoregulation are not uncommon in preterm infants; however, a correlation between the occurrence of impaired autoregulations and brain lesions was not observed.3–5
Using cerebral blood oxygenation to identify critical CBF thresholds is potentially confounded by the indirect relationship between cerebral blood oxygenation and CBF, as the former also depends on the cerebral metabolic rate of oxygen (), the cerebral blood volume (CBV), and arterial oxygen saturation.6 An alternative approach would be to combine NIRS with diffuse correlation spectroscopy (DCS), an emerging optical method sensitive to CBF.7–9 This combination has the advantage of providing a means of monitoring both CBF and .10–12 The latter can be determined from blood oxygenation and flow measurements and is considered a more sensitive indicator of tissue viability.13,14 For example, has been shown to be more sensitive to the severity of cerebral hypoxia-ischemia than NIRS oxygenation measurements alone.15,16 The combination of NIRS and DCS could therefore help identify clinically significant passive-pressure CBF by detecting flow fluctuations large enough to affect cerebral energy metabolism.
The purpose of this study was to demonstrate that changes in absolute could be measured by combining NIRS and DCS. To quantify , techniques for measuring absolute CBF and cerebral blood oxygenation are required. In this study, relative blood flow changes measured by DCS were converted into units of CBF using a bolus-tracking time-resolved (TR) NIRS technique to measure baseline CBF.17 Cerebral blood oxygenation was determined from multiwavelength TR-NIRS measurements. The accuracy of the oxygenation measurements was assessed by directly measuring blood oxygenation in the sagittal sinus. The sensitivity of the hybrid approach to changes in was investigated by manipulating cerebral metabolism in newborn piglets by altering the anesthetics and by injecting sodium cyanide, a mitochondrial inhibitor.18,19
This study was approved by the Animal Use Subcommittee at Western University. Experiments were carried out on newborn Duroc pigs ( days old). All surgical procedures were performed while piglets inhaled 3% to 4% isoflurane. Piglets were tracheotimized and mechanically ventilated on an oxygen/medical air mixture. Catheters were inserted into an ear vein, the left femoral artery, and superior sagittal sinus through a burr hole drilled into the skull. The ear catheter was used to inject the light-absorbing dye indocyanine green (ICG) (Sigma-Aldrich) and the different drugs used in the experiment. The femoral and sagittal sinus catheters were used to acquire blood samples to determine the oxygen saturation of arterial blood () and cerebral venous blood (), respectively. The superior sagittal sinus primarily drains the cerebral cortex, which is the brain region interrogated by the optical probes.20 The femoral line was also used to monitor heart rate, blood pressure, blood gases ( and ), and blood glucose. A 1- to 2-mL infusion of a 25% glucose solution was administered intravenously if glucose levels fell below . A heated water mattress was used to maintain rectal temperature between 37.5°C and 38.5°C throughout the experiment. After surgery, the optical probes were positioned on the scalp using a custom-made probe holder with a source-detector separation of 20 mm. A separation of 10 mm between the probes from the two optical systems ensured that they interrogated roughly the same brain region. No data were acquired for at least 30 min after surgery to ensure the piglet was physiologically stabile. Baseline conditions were identified by blood samples revealing normal and : 38–42 and 100–170 mmHg, respectively.21
Five different cerebral metabolic states were induced by manipulating the anesthetics and by injecting sodium cyanide. After a change of condition, data acquisition was delayed by approximately 5 min to allow time for cerebral metabolism to stabilize. Under each condition, DCS was used to measure the blood flow index (BFI), and multiwavelength TR-NIRS was used to measure the cerebral tissue oxygen saturation () (see instrumentation section). Blood samples (0.3 to 0.5 mL) were drawn from the femoral artery and the sagittal sinus to determine and , by hemoximetry (ABL80 Flex Co-ox, Radiometer). The DCS and TR-NIRS data were acquired in three blocks during each condition (Fig. 1). In each block, 96 TR-NIRS measurements were acquired with an integration time of 1 s, and 20 DCS measurements were acquired with an integration time of 30 s. All data sets for each condition were acquired within approximately 45 min.
Initial baseline measurements were acquired under 1.75% to 2% isoflurane mixed with a combination of medical air and oxygen to maintain normal . In addition to the standard data acquisition protocol, CBF was measured by a dynamic contrast-enhanced TR-NIR method using ICG as an intravascular contrast agent.22 This method requires injecting an intravenous bolus of ICG () and measuring the time-varying concentration of ICG in arterial blood and the brain. The brain ICG concentration curve was determined by acquiring a series of TR-NIR measurements at a sampling rate of 400 ms. The arterial concentration curve was measured using a pulse dye densitometer (DDG 2001, Nihon Kohden) attached to a foot. Three successive sets of ICG data were acquired to improve the precision of the baseline CBF measurement. Successive ICG injections were separated by 10 min to allow time for ICG clearance from the previous injection. The average value of CBF was used to convert subsequent DCS BFI measurements into units of CBF.17
After the baseline measurements, TR-NIRS and DCS were used to measure and the BFI, respectively, under four conditions. First, was increased by discontinuing isoflurane and starting an intravenous infusion of fentanyl () combined with inhalation of 70% nitrous oxide (). Second, the barbiturate pentobarbital () was injected intravenously to decrease . Cerebral energy metabolism was further reduced by switching off the mixture, returning the piglet to isoflurane (1.75% to 2%), and administering another of pentobarbital. For the final condition, sodium cyanide was injected intravenously at a dose of . This nonlethal dosage of cyanide is insufficient to completely inhibit mitochondrial function;23 however, it is sufficient to reduce mitochondrial respiration.19
Time-resolved NIRS system
The TR instrument for the measurements consisted of three picosecond diode lasers emitting at 760, 802, and 830 nm (LDH-P-C-810, PicoQuant). The output and pulse repetition rate of each laser were set to 1.4 mW and 29.3 MHz, respectively. A variable neutral density filter (NDC-50-4M, Thorlabs) was placed in front of each laser to adjust the intensity of the beam before the light was coupled into a multimode fiber (emission probe: , , 4.7 mm outer diameter; Fiberoptics Technology). Fibers of different lengths, forming a bundle, were used to provide time multiplexing.24 For the ICG bolus-tracking measurements, only the 802-nm laser was used at a repetition rate of 80 MHz to optimize the signal-to-noise ratio (SNR).25
Reflected light from the head was collected with a 1.5-m-long multimode fiber placed 20 mm from the emission fiber bundle on the piglet’s intact head. The collected photons were then sent to a Peltier-cooled photomultiplier tube (PMT) (PMC-100, Becker & Hickl) coupled to a time-correlated single photon counting module (SPC-134, Becker & Hickl). Temporal point spread functions (TPSFs) were computed by synchronizing the photon detection with the laser pulse trigger provided by the driver (PDL 828, PicoQuant). Each TPSF was acquired for 1 s for the oxygenation measurements and 400 ms for the ICG bolus-tracking method. Temporal dispersion caused by the system was corrected for by measuring the instrument response function (IRF).26 To avoid artifacts such as instrument temporal drift, the TR-NIRS system was allowed a 1.5-h warm-up delay period prior to the experiment.25,27
Diffuse correlation spectroscopy
The DCS light source was a continuous-wave laser emitting at 785 nm (DL785-100-S, CrystalLaser, Nevada) with a maximum output power of 100 mW and a coherence length . Similar to the time-resolved setup, the laser beam was first attenuated by a variable neutral density filter (NDC-50-4M, Thorlabs, New Jersey) and then coupled into an emission fiber (emission probe: , , 4.7 mm outer diameter; Fiberoptics Technology, Connecticut). Photons scattered from tissue were detected at a distance of 20 mm from the emission probe using a 4-m single-mode fiber (, , , single-mode cutoff wavelength at 1260 nm). Because the detection fiber is a few-mode fiber at the laser emission wavelength, the fiber was wrapped into a 15-cm coil to attenuate the higher-order modes by converting them into nonpropagating modes.17,28,29 Photons were detected by a single photon counting module (SPCM-AQR-15-FC, PerkinElmer Canada Inc, Quebec, California). The output from the detector was sent to a correlator board (DPC-230, Becker & Hickl, Berlin, DE) to compute the normalized intensity autocorrelation function.
Measuring cerebral oxygen saturation by TR-NIRS
Brain tissue optical properties (i.e., the absorption and reduced scattering coefficients, and , respectively) were quantified using the solution to the diffusion approximation for a semi-infinite turbid medium with extended boundary conditions.30 The three sets of TPSFs collected at each condition were averaged together and were fitted by the theoretical model convolved with the measured IRF.27 Initial values for , , and an amplitude scaling factor were obtained by analyzing the baseline data with a three-parameter nonlinear fitting routine.26 The amplitude term was included in the fitting to take into account variations in laser power, detection gain, and coupling efficiency.26,27 To improve the stability of the fitting algorithm, the TPSFs were analyzed for each metabolic condition using only as a fitting parameter; and the amplitude scaling factor were fixed to the values retrieved from the three-parameter fit at baseline. This approach was reasonable since the metabolic changes affect only blood oxygenation and not tissue scattering properties. However, in more extreme conditions that could also alter light scattering, such as large changes in the total hemoglobin concentration,31 could be included as a fitting parameter. The concentrations of oxy- and deoxy-hemoglobin ( and [Hb], respectively) were derived from the measured values using the known wavelength-dependent extinction coefficients and assuming a cerebral water content of 85% for neonatal pig.32 Cerebral oxygen saturation, , was defined as
Measuring absolute CBF by TR-NIRS
The methodology underlying the TR-NIRS bolus-tracking technique for quantifying CBF, which is used extensively with imaging modalities such as magnetic resonance imaging and computed tomography,33 assumes that the microvasculature can be modeled as a linear time-invariant system. For this application, linearity means that the contrast agent concentration in brain tissue is linearly proportional to the concentration in arterial blood, and time-invariance implies that the hemodynamic properties must remain constant during the acquisition period. Under these assumptions, the time-varying concentration of ICG in tissue, , is related to the arterial blood ICG concentration curve, , by the convolution operator34 is determined from change in due to the passage of ICG through the cerebral microvasculature, and was measured by the DDG as outlined in Sec. 2.2.26 The flow-scaled impulse residue function, , was extracted from and by performing a deconvolution and its initial height equals , since by definition .22
Measuring changes in CBF and by TR-NIRS/DCS
The DCS data from each cerebral metabolic condition were analyzed by the solution to the correlation diffusion equation for a semi-infinite homogeneous medium assuming Brownian motion of scatterers.7,8,35 The model was fitted to the normalized intensity autocorrelation function () using the and values obtained by TR-NIRS. The fitting parameters were a scaled diffusion coefficient, which is referred to as the BFI, and a coherence factor ().17 The change in the BFI at a given condition relative to baseline, denoted rCBF, is given by36 37 and [tHb] is the total hemoglobin concentration. In these experiments, was measured by pulse oximetry, [tHb] was determined from the average of two baseline arterial blood samples, and CBF was determined from the calibrated DCS blood flow measurements. The remaining parameter was determined directly from the sagittal-sinus blood samples and also from the TR-NIRS measurements. Equation (5) neglects the contribution of dissolved oxygen in plasma, which is reasonable at normal values.38,39 For TR-NIRS, was determined from by assuming the cerebral blood volume is comprised of a known venous volume fraction (), since represents the average cerebral blood oxygenation: 4041.42.–43), and using Eq. (6), the Fick principle for the optical measurements is given by
Statistical analyses were conducted using SPSS 20.0 (SPSS, Chicago, Illinois). A repeated-measures analysis of variance (ANOVA) was used to identify significant differences in and between metabolic conditions and between the NIRS and blood-sampling methods. Bland–Altman analysis was used to compare corresponding measurements from the two methods.44 Post hoc analyses were conducted to identify significant differences at each condition, with respect to baseline values, for all measured parameters. Statistical significance was defined as , and all data are presented as error of the mean (SEM) unless otherwise noted. In one animal, measurements were not acquired during the condition, which was corrected for by using the missing value analysis (MVA) regression algorithm.45
A Monte Carlo–type approach was conducted to investigate how errors in the baseline parameters would affect subsequent measurements from the one-parameter fitting routine. First, a theoretical TPSF was generated using the solution to the diffusion approximation for a semi-infinite homogeneous medium and a set of typical experimental values of the scaling amplitude, and (8000, 0.255, and , respectively).30 The simulated TPSF was convolved with an experimental IRF and Poisson noise added to reflect typical experimental data. The noisy TPSF was analyzed by the same nonlinear fitting routine used with the experimental data to generate best-fit estimates of the baseline parameters. Next, another noisy TPSF was generated using the initial input values, and a one-parameter fit performed to extract an estimate . In this step, the values of the amplitude factor and were set to the estimates from three-parameter fit. The entire procedure was repeated 5000 times to generate a distribution of best-fit values for each fitting parameter.
To determine the error in due to uncertainties in the two parameters measured under each condition (i.e., BFI and ), the SEM of each parameter was determined from the series of measurements acquired at baseline (i.e., 60 and 288 measurements for BFI and , respectively). A repeated-measures ANOVA was used to determine the precision of the baseline CBF values measured by the TR-NIRS bolus-tracking method.
A total of 12 piglets were studied; however, three were excluded because the sagittal sinus was inadvertently punctured during insertion of the catheter. Puncture caused blood to leak into the cerebral spinal fluid, resulting in erroneous NIRS measurements of  and [Hb].18 One additional experiment was excluded due to technical errors with the TR-NIRS system, which results in incomplete TPSFs. The average physiological parameters at each metabolic condition from the eight successful experiments (three male, five female, mean weight , mean age days) are given in Table 1. Of all the measured parameters, only showed no significant differences between conditions. Blood pH and rectal temperature were significantly different after injecting sodium cyanide owing to the systemic effect of the drug on metabolism. Significant differences in were found under the two conditions involving , which can be explained by a reduction in the inhaled oxygen fraction during 70% inhalation. Both HR and MAP showed general changes with the different anesthetics as expected.36 The average [tHb] measured under baseline conditions from the hemoximeter was . The tissue optical properties measured under baseline conditions at 760, 802, and 830 nm were , , and for , respectively, and , , and for .
Measured physiological parameters.
Figure 2 shows a sample of DCS intensity autocorrelation curves collected from each induced metabolic condition during a single experiment. A steeper DCS decay curve is represented by a larger BFI value, indicating faster blood flow. Typical arterial and brain ICG concentration curves obtained with the DDG and TR-NIRS, respectively, are shown in Fig. 3. The derived baseline CBF value obtained from the ICG data sets was used to calibrate all subsequent BFI values obtained by DCS. The measured CBF values obtained from the dynamic contrast-enhanced technique ranged from 19 to with an average of .
Significant differences were observed in all the BFI values listed in Table 2 compared to baseline, except when pentobarbital was injected while under anesthesia. No statistical significant differences were observed between conditions for the retrieved baseline value of , which was also in agreement with the measured value obtained by Diop et al.17 No significant differences were observed in at any of the conditions. A repeated-measures ANOVA showed a significant overall effect by condition for [, , ], but there was no significant effect by technique [, ]. Similarly, there was a significant effect by condition for [, , ], but no significant effect by technique [, ]. These results indicate that the various conditions altered cerebral energy metabolism, but there were no significant differences in the and measurements between the sagittal sinus blood measurements and the DCS/NIRS method.
NIRS and blood sample measurements.
Figure 4 illustrates the average and measurements obtained from the blood sample and DCS/TR-NIRS techniques under the different conditions. Measurements were obtained over a range from 25% to 70% for and from 0.3 to 4 mL for . A Bland–Altman plot of the difference between individual measurements from the two techniques is shown in Fig. 5. The mean difference was 0.027 mL , with limits of agreement (i.e., 95% boundaries) of 0.861 and .
Figure 6 shows the predicted relationship between errors in the baseline optical properties to the errors in the amplitude factor as determined from the Monte Carlo simulations. The figure demonstrates that at typical experimental noise levels, and were relatively insensitive to errors in the amplitude factor. For example, the error in either or was only 6% in the extreme case of a 30% error in the amplitude factor. From the Monte Carlo simulations, the coefficient of variation of for the one-parameter fitting routine was 1.7%. The small magnitude of this error was also reflected in the SEM of (0.1%) determined from the series of baseline measurements. Similarly, the SEM of the baseline BFI was 2%. The repeat measurements of absolute CBF from the bolus tracking method indicated that this was the greatest source of error, with an estimated precision of 11.7%.
The main finding of this study was that changes in absolute could be measured by combining two near-infrared techniques: DCS and TR-NIRS. This was accomplished by combining quantitative CBF and measurements obtained by a previously validated DCS/TR-NIRS approach and multiwavelength TR-NIRS, respectively. The accuracy of the measurements was verified by comparison to direct measurements of cerebral blood oxygenation obtained from sagittal sinus blood samples. CBF, , and were measured over a metabolic range from approximately 0.3 to 4 mL with no evidence of any level-dependent bias in either the or measurements (Figs. 4 and 5). DCS and NIRS have been combined previously to measure relative changes in ;9–12,46 however, this is the first study to use this combination to quantify . The success of this combined optical approach ultimately depends on the accuracy of the underlying CBF and measurements. A number of studies have demonstrated the ability of DCS to track relative CBF.17,47–50 In particular, Zhou et al.49 reported a strong correlation between changes in CBF measured by DCS and those measured by fluorescent microspheres in newborn piglets. Similarly, we previously reported a strong correlation between CBF measurements obtained by the bolus-tracking TR-NIRS technique and the BFI determined from DCS ().17
In our previous study, a single pulsed laser was used to measure the absorption changes caused by the flow of ICG through brain tissue.22 Determining  and [Hb], and thereby , requires measuring the tissue optical properties at multiple wavelengths. For this purpose, the TR system was expanded to three pulsed lasers, and time multiplexing was used to acquire the corresponding three TPSFs within the same time window.24 The disadvantage of this approach is a lower count rate per channel due to the reduced pulse repetition rate (29.3 compared to 80 MHz); however, for steady-state oxygenation measurements, this was not a concern, as the SNR can be improved by increasing the integration time. To quantify the wavelength-dependent and values, the IRF was measured for each laser at similar count rates to those obtained experimentally. Temporal drift and variations in the laser pulse width (i.e., jitter) are potential sources of error when measuring tissue optical properties by TR-NIRS.27 To minimize the former, every experiment was preceded by a 1.5-h delay to allow sufficient time for the TR-NIRS system to stabilize to avoid instrument drift during the course of an experiment.25 The results displayed in Fig. 6 indicate that the baseline and values could be determined with reasonable accuracy even if there were large fluctuations in the amplitude factor, such as due to jitter. A possible improvement to the system would be to correct for both drift and jitter by acquiring reference data throughout the experiment.51
The precision of will depend on the uncertainties in the baseline optical properties, CBF and . The Monte Carlo simulations demonstrated that the uncertainties in the optical properties were small for noise levels typical of the experimental TPSFs and had little effect on subsequent values determined from the one-parameter fitting routine. These predictions were in good agreement with the SEM of (0.1%) determined from the series of TPSFs acquired during the 5-min baseline period. Errors in CBF determined by the calibrated DCS approach will depend on the precision of both the baseline CBF determined by the bolus-tracking method and the relative BFI obtained by DCS. From repeat bolus injections at baseline, the precision of the former was estimated to be 11.7%, which is similar to the value determined previously (9.7%) with a continuous-wave NIRS system.22 In these experiments, the SEM of the BFI was due to the large number of acquisitions obtained in each condition. A variability of would still have been achieved if the acquisition time had been reduced to 5 min (i.e., 10 measurements). Maintaining the same precision at shorter acquisitions would require increasing the number of detectors,28,47 as only a single SPCM was used to acquire the DCS data in these experiments. In addition to the uncertainties in the measurements parameters, the accuracy of the measurements will depend on the assumed value of the cerebral venous blood volume fraction (). The need to assume a value is a common limitation with measurements obtained by NIRS and positron emission tomography.18,39,52 The good agreement between the NIRS measurements and the values derived from sagittal-sinus blood samples (Fig. 5) indicate that the assumed value of 75% is reasonable. However, any variation in across animals or subjects will clearly reduce the accuracy of the measurements. It may be possible to circumvent this potential source of error by using NIRS spiroximetry to measure venous oxygen saturation.53
The retrieved values at baseline were similar to those of Diop et al.17 and Ijichi et al.43 However, the average at 802 nm was greater than the values given in the two previous studies: (Ref. 17) and (Ref. 43). The cause of this difference is uncertain; however, the average value () was similar to baseline values previously reported in piglets.54–57 In addition, the values determined from the absorption changes at the different metabolic conditions ranging from 25% to 70% were in good agreement with the corresponding values measured directly from sagittal sinus blood samples (Table 2, Fig. 4). Both sets displayed the same trend in venous oxygenation, as cerebral metabolism was initially increased under anesthesia and subsequently decreased with successive injections of pentobarbital.
The overall agreement between measurements derived from TR-NIRS and sagittal sinus values is similar to our previous validation study, which was conducted using newborn piglets but involved a broadband CW-NIRS system.18 In that study, the mean difference between the two techniques was 0.006 mL , and the 95% confidence interval was . The latter is slightly smaller than the boundaries shown in Fig. 5. However, this difference can be explained by the outlier that had an average difference greater than the 95% confidence interval. Removing this point reduced the range to , in excellent agreement with the previous results. It is interesting that the overall trends reported in the two studies were in good agreement considering the different approaches used to quantify cerebral blood oxygenation. The CW-NIRS spectral data from the previous study were characterized by second derivative spectroscopy, which can only estimate [Hb] and not , since the latter has no definitive features in the derivative spectrum.58 Consequently, was determined by normalizing [Hb] by the cerebral blood volume. This step is not required for TR-NIRS, since both [Hb] and  are determined from the measured optical properties. A further advantage to directly measuring and by TR-NIRS is that these values can also be used in the analysis of DCS data rather than assuming known values. Another advantage with TR-NIRS is its superior depth sensitivity, which could prove useful when adapting these optical technologies to measuring CBF and in adult patients.59,60
Although the two studies demonstrate similar agreements between the NIRS and sagittal sinus measurements, there was a large difference in the initial baseline values: in the current study and from Tichauer et al.18 One possible explanation is the difference in baseline anesthetics in the two studies. Baseline measurements in the current study were acquired using a higher concentration of isoflurane (1.75% to 2% compared to 1.5%), which will reduce cerebral energy metabolism. Another contributing factor could be the difference in CBF measurements from CW- and TR-NIRS techniques. Previously, we have demonstrated that the CBF values from TR-NIRS underestimated the CW-NIRS values, although the reason for this difference was unclear.26 Adjusting the current values for this difference would increase the mean baseline value to , which is still below our previous results. Most likely, the overall difference between the two studies is a combination of both factors.
In addition to the different anesthetics, measurements were also conducted after administering sodium cyanide. This drug was used to alter cerebral oxidative metabolism by an alternative means, namely, the inhibition of mitochondrial cytochrome oxidase. Overall, the agreement between the DCS/TR-NIRS and sagittal sinus values under this condition was similar to that observed under the different anesthetics. However, the metabolic effect of cyanide was smaller than expected based on a previous study involving newborn piglets.19 This muted effect is likely a result of injecting the sodium cyanide while the animals were under the deepest anesthetic (pentobarbital plus isoflurane). Nevertheless, cyanide did cause a small reduction in , which was observable by both techniques.
A potential limitation with the current study is that the CBF measurements obtained by the calibrated DCS technique were used to convert both the sagittal-sinus and NIRS venous oxygenation measurements into . Consequently, any systemic error in the optical CBF measurements would not be observed in this study. We previously demonstrated a very strong correlation between relative flow changes measured by DCS and CBF values measured by TR-NIRS with a slope from the linear regression of 1.05 (Ref. 17). This comparison was conducted over a range of CBF values from 17 to , which is very similar to the range observed in the current study. However, more validation studies would be required if the combined DCS/TR-NIRS approach were used to measure under pathological conditions such as during ischemia or severe hypoxia that could result in greater changes in either CBF or .16,43
In summary, this study demonstrated that the combination of DCS and TR-NIRS can be used to measure as verified by the comparison to values derived from cerebral venous blood samples. Experiments were conducted using piglets because they are similar in size to human newborns and, therefore, the DCS/TR-NIRS system could be used in clinical studies. Adapting this approach to the neonatal intensive care unit will require synchronizing the two optical systems to provide truly continuous CBF and measurements. Extending this approach to adult patients represents a significant challenge, as it requires depth-resolved techniques, such as the use of multilayered modeling approaches, to accurately measure CBF, cerebral oxygenation, and BFI.61–63 Incorporating sensitivity functions for the different tissue layers into the analysis of bolus tracking data has been shown to improve the accuracy of CBF measurements in animal models; however, clinical studies will require validating this approach in human subjects.60,64
We acknowledge the Canadian Institutes of Health Research and Natural Sciences and Engineering Research Council for financial support. We also thank the animal technicians, Jennifer Hadway and Lise Desjardins, for their assistances and expertise with the experiments and Dr. Yves Bureau for his advice on the statistical analysis.