Open Access
Add to BibSonomy
Abstract
Near-infrared spectroscopy (NIRS) is considered ideal for brain monitoring during preterm infancy because it is non-invasive and provides a continuous measure of tissue oxygen saturation (StO2). Hyperspectral NIRS (HS NIRS) is an inexpensive, quantitative modality that can measure tissue optical properties and oxygen saturation (StO2) by differential spectroscopy. In this study, experiments were conducted using newborn piglets to measure StO2 across a range of oxygenation levels from hyperoxia to hypoxia by HS and time-resolved (TR) NIRS for validation. A strong correlation between StO2 measurements from the two techniques was observed (R2 = 0.98, average slope of 1.02 ± 0.28); however, the HS-NIRS estimates were significantly higher than the corresponding TR-NIRS values. These regression results indicate that HS NIRS could become a clinically feasible method for monitoring StO2 in preterm infants.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Near-infrared spectroscopy (NIRS) is becoming recognized as a cost-effective, non-invasive monitor of cerebral hemodynamics, particularly for neonates because of their relatively thin skull and scalp [1]. The majority of commercially available NIRS systems monitor tissue oxygen saturation (StO2) by measuring light absorption at a few discrete wavelengths [2]. In comparison, hyperspectral NIRS (HS NIRS) that utilizes a broadband light source and a spectrometer to capture the entire attenuation spectrum offers some unique advantages. Notably, the differential path-length factor (DPF) can be estimated by combining second derivative spectroscopy with the modified Beer-Lambert law [3], and used to determine the concentration of deoxyhemoglobin (Hb) in the neonatal brain [4]. Knowing DPF also provides a means of measuring cerebral blood flow (CBF) using indocyanine green (ICG) as an intravascular contrast agent [5]. Measurements of CBF and Hb concentration can be combined to estimate the cerebral metabolic rate of oxygen (CMRO2) [6] [7]. Importantly, HS NIRS also provides the most robust approach for monitoring changes in cytochrome-c-oxidase (CCO) [8,9].
A limitation with analyzing attenuation spectra using the modified Beer-Lambert law is that the second derivative of the oxyhemoglobin (HbO2) spectrum does not have resolvable features in the near infrared range, and consequently, StO2 cannot be determined. One solution is to include both first and second derivative spectra since the former will contain HbO2 contributions. This modification requires replacing the modified Beer-Lambert law in the spectral analysis with a theoretical model of light propagation in tissue such as the diffusion approximation to account for light scattering [10,11]. With this approach, water and Hb concentrations are determined from the second derivative spectrum, and HbO2 from the first derivative spectrum by modeling the wavelength dependency of scattering by a power law [12]. By combining StO2 measurements with CBF estimates, HS NIRS has been used to assess the effects of clinical intervention on CBF and CMRO2 in preterm infants with patent ductus arteriosus and post-hemorrhagic ventricular dilatation [13,14]. More recently it was combined with diffuse correlation spectroscopy to provide continuous monitoring of CBF, CCO, and StO2 [15].
Despite its applications in clinical studies, the sensitivity of HS NIRS to oxygenation reductions has not been demonstrated, nor have the StO2 estimates been compared to those from an another NIRS method. In the current study, these capabilities were assessed in a piglet model of hypoxia that involved reducing the fraction of inspired oxygen (FiO2) in a stepwise manner [16]. Spectra were acquired at each oxygenation level, and for validation, multi-wavelength time-resolved (TR) data were acquired for independent measurements of the absorption and reduced scattering coefficients (μa and μs′, respectively). StO2 obtained by multi-wavelength TR NIRS were previously shown to be a good agreement with direct measurements of venous blood oxygenation from the sagittal sinus [17]. In addition to these experiments, Monte Carlo simulations were performed across a wide range of StO2 values to investigate the precision and accuracy of the fitting method as a function of spectral signal-to-noise ratio (SNR).
2. Methods
2.1 Instrumentation
The light source of the HS-NIRS system is a 20-W halogen lamp (Ocean Optics HI-200-HP) coupled to an optical fiber bundle (3.5 mm active diameter, 30 µm core fibers, 0.55 numerical aperture) to direct the light to the scalp. Diffusely reflected light was collected by another fiber bundle (3.85 mm active diameter, 50 µm core, NA = 0.22) at a distance of 3 cm from the source and directed to a custom-made spectrometer (P&P Optica, ON, Canada). The spectrometer incorporated a 1024×256 pixel CCD Camera that has a resolution of 1.65 nm and was operated at −80°C, using a spectral range from 548 to 1085 nm, and an exposure time of 250 ms (Andor iDUS 420 BEX2-DD, Oxford Instruments).
The in-house-developed TR-NIRS system included four picosecond pulsed diode lasers operating at wavelengths of 670, 760, 804 and 830 nm (LDH-P-C PicoQuant, Germany). The repetition rate was set to 40 MHz using a computer-controlled laser driver [18,19] (PDL 828, PicoQuant). These lasers were coupled to two bifurcated fibers (active diameter = 0.4 mm, NA = 0.39, Thor labs, United States) that directed the light from two lasers to a common end point on the scalp. Diffusely reflected light was collected by another fiber bundle (active diameter = 3.6 mm, NA = 0.55, Fiberoptics Technology) set at a distance of 3 cm from the emission fiber and coupled to a hybrid photomultiplier tubes (PMA Hybrid 50, PicoQuant). A HydraHarp 400 (PicoQuant) time-correlated single-photon counting unit was used to record the arrival times of the photons. A distribution of times of fight (DTOF) was generated using LabView Software (National Instruments, TX, United States) [20]. Finally, the instrument response function (IRF) for each wavelength was acquired using a 6-cm long light-tight box constructed of polyvinyl chloride. The box contained a neutral density filter (Thor Labs, NJ, United States) to reduce light intensity and a thin piece of white paper to disperse the light beam [21]. The temporal delay due to light traveling the length of the IRF box was accounted for in the analysis.
2.2 Animal preparation
All animal experiments were conducted in accordance with the guidelines of the Canadian Council of Animal Care (CCAC) and approved by the Animal Care Committee at Western University. Newborn piglets were anesthetized at induction with 5% isoflurane, which was reduced to 3% for surgical procedures, and maintained at 2-2.5% isoflurane throughout the experiments. Animals were tracheotomized and mechanically ventilated on a mixture of 2 L oxygen and 2 L medical air. Vital signs, including heart rate (HR), arterial O2 saturation (SaO2), end-tidal CO2 tension, respiratory rate and temperature were monitored (SurgiVet, Smith Medical, MN). Arterial blood pressure (MAP) was continuous monitoring via a femoral artery catheter. Arterial blood samples were intermittently collected via the femoral catheter for blood gas and glucose analysis (Radiometer, ABL 80 FLEX CO-OX, Denmark). Lateral ear and cephalic veins were catheterized for IV access and infusion of NaCl fluids (4 ml/kg/h) for the duration of the study. The top of the head was shaved, and the piglet was placed in prone position on a plexiglass jig to help support a 3D-printed probe holder fastened to the piglet’s head, a schematic of the probe holder configuration is depicted below. The probe holder was printed with a slight curvature to approximate the shape of the piglet head. It was printed using a semi-flexible material that could bend to ensure good probe contact but sturdy enough to support the fibers (Fig. 1).
2.3 Experimental procedure
To achieve stepwise changes in oxygenation, FiO2 was varied by manipulating the proportions of oxygen, medical air (21% O2, balance N2) and nitrogen delivered through the mechanical ventilator. After each reduction of FiO2, stable SaO2, HR and MAP measurements were used to confirm that the animal had reached a physiological steady-state. Once confirmed, an arterial blood sample was taken to record arterial oxygen and carbon dioxide tensions (PaO2 and PaCO2, respectively), total hemoglobin concentration and blood glucose. NIRS measurements were then acquired sequentially: time-resolved first, followed by hyperspectral. A final blood sample was acquired to confirm that the blood gases had remained stable during data acquisition. Each experiment was comprised of seven-to-ten oxygenation stages, starting with hyperoxia, followed by a serial reduction in O2 to achieve varying levels of hypoxia. The variability in the responses to FiO2 reductions across piglets dictated the number of steps and the minimum FiO2 that could be achieved. The lower limit of oxygenation was determined by the veterinary technician by evaluating each piglet’s HR and MAP responses to lower FiO2 levels.
TR-NIRS measurements consisted of two sets of data, each comprised of 256 averages acquired over 86 s to obtain DTOFs with high SNR. For each data set, DTOFs were acquired for two wavelengths: first 670 and 804 nm, followed by 760 and 830 nm. The wavelengths were separated in sets of two to avoid temporal overlap (i.e., crosstalk) between DTOFs collected within a 25-ns time window. A second blood draw was taken for gas analysis prior to acquiring the HS-NIRS data to ensure systemic parameters remained stable. The HS-NIRS protocol consisted of collecting 256 spectra in a 64-s period to obtain high SNR spectrum.
2.4 Data analysis
The analysis of the spectral data was performed using the solution to the diffusion approximation for a semi-infinite medium [22]. This approach is acceptable because signal contamination from the extra-cerebral layer is small at a source-detector separation of 3 cm due to a relatively thin scalp and skull of the young piglets [23]. The absorption coefficient was defined as a sum of the three main endogenous chromophores: Hb, HbO2, and water:
Spectral analysis began by applying a wavelet de-noising algorithm that first transformed the noise from Poisson to Gaussian noise using the Anscombe transformation on the acquired spectra. This Gaussian noise was then removed by wavelet de-noising, and applying the inverse Anscombe transformation to obtain noise-reduced spectra [25]. Smoothing was performed using a 5-point moving average filter before generating derivative spectra to minimize noise contributions. Chromophore concentrations and the scattering terms were determined in a series of steps involving fitting the numerical derivatives of the theoretical model (i.e., the analytical solution to the diffusion approximation) to the derivative spectra. First, WF was determined by fitting the water feature between 815 and 840 nm in the second derivative spectrum. Using the determined WF, the Hb concentration was obtained by fitting the second derivative spectrum between 690 to 775 nm to capture the characteristic Hb feature centred at 760 nm. With these two parameters fixed, the first derivative of the model was fit to the first derivative spectrum to determine the HbO2 concentration and the scattering terms β and α between 680 to 875 nm. Fitting was performed using a constrained optimization routine based on the MATLAB script fminsearchbnd with upper and lower boundaries set to span published values and are displayed in Table 1 [12].
Table 1. Baseline and boundary values for HS-NIRS fitting
For each animal, the scattering terms and the water fraction were included as fitting parameters in the analysis of the spectrum acquired at the baseline oxygenation (60% O2). For subsequent steps, these parameters were fixed to their initial values since altering FiO2 is not expected to affect tissue scattering properties or brain water content. At each oxygenation level, the estimates of HbO2 and Hb concentrations were used to compute StO2 according to:
The TR-NIRS data were analyzed using the time-dependent solution to the diffusion approximation for a semi-infinite medium [22]. The first step was to convolve the model solution with the measured IRF to correct for temporal dispersion caused by the system. Next, the model was fit to the acquired DTOFs using three parameters: μa and μs′, and an amplitude factor that accounts for laser power, detection gain and coupling efficiency [26]. Fitting was performed using an optimization routine based on the MATLAB script fminsearch. The fitting range was set to 80% of the peak value on the ascending edge and 20% on the descending edge of the DTOF [21,27]. Similar to the analysis of the HS-NIRS data, all three parameters were determined under the initial hyperoxia condition, therafter μs′ and the amplitude factor were fixed to their initial values for the analysis of all subsequent measurements. Maintaining the same technical set-up throughout each experiment ensured that the amplitude term would not change between FiO2 levels.2.5 Error analysis
Monte Carlo simulations were conducted to assess the accuracy of the 3-step, multi-parameter fitting approach outlined in section 2.4. Simulated spectra were generated using the semi-infinite solution to the diffusion approximation with values of Hb and HbO2 that corresponded to StO2 values of 70, 50 and 35% (all input parameters are given in Table 2). Poisson noise was added at each wavelength, and the same fitting routine used to analyze the piglet data was used to obtain best-fit estimates of the 5 parameters with the same initial values and boundary conditions given in Table 1. Simulations were repeated 500 times to generate a distribution of estimates for each fitting parameter. The entire procedure was conducted by averaging spectra over 16 repetitions, which corresponds to an acquisition time of 4 s. In addition, the procedure was repeated for StO2 = 70% for averages ranging from 1 to 256 frames to estimate the precision over a range of SNRs.
Table 2. Input parameters for the Monte Carlo simulations at StO2 = 70, 50 and 35%
2.6 Statistical analysis
SPSS 25 (IBM, NY, United States) was used for statistical analyses involving ANOVAs and mixed linear models, while Excel was used for more basic statistics such as t-tests. Comparisons of StO2 and µa measurements between the two NIRS techniques at different FiO2 levels were investigated by a two-factor repeated-measures ANOVA with level of FiO2 and technique being the factors. Standard linear regression was not used to assess the significance of correlation between the two sets of StO2 measurements since multiple measurements were collected from each animal. Instead, a variation of the generalized estimating equation technique was utilized. First, a linear fit was applied to data from each piglet individually. Second, a significant correlation was tested by using a t-test to compare the average of the distribution of slopes against the null hypothesis (i.e., slope = 0). Finally, the distribution of slopes was compared to a slope of 1 to determine the agreement with the line of identity. A t-test was used to compare changes in physiological parameters to the baseline value to compare the μs′ measured by the two NIRS modalities.
A linear mixed effects model was used to estimate inter-subject variability with StO2 as the random variable and subject as a factor. In the mixed model analysis, the coefficient of variation was used as a measure of variability in the StO2 values across the common FiO2 conditions experienced by all subjects. Finally, a Bland Altman plot was constructed to further compare StO2 measured by the two NIRS systems in order to investigate potential biases between the techniques. All values are presented as means ± standard deviation, F-statistics were presented with the associated degrees of freedom, and statistical significance was defined as p < 0.05.
3. Results
3.1 Physiological parameters
Eight piglets (4 females and 4 males measured, one female was removed due to a lack of physiological response to changes in FiO2.) were studied with an average age of 4.4 ± 2.3 days and weight of 2.3 ± 0.7 kg. Data were successfully acquired at 7 to 10 oxygenation levels in each piglet. The number of levels varied based on the animal’s HR, MAP and SaO2 response to decreasing FiO2 levels as stated in section 2.3. Table 3 presents a summary of the physiological parameters at six FiO2 states. These six states were selected as they spanned a wide range of oxygenation and were common across all animals. As expected, both SaO2 and PaO2 dropped significantly as FiO2 was reduced. No statistically significant differences across the different FiO2 levels were found for PaCO2, MAP, and temperature. However, HR rose significantly at the most extreme hypoxic condition. In total, 68 sets of TR-NIRS and HS-NIRS measurements were completed. The mean difference in PaO2 measurements acquired under the same FiO2 state was −0.31 ± 4.4%, confirming that steady state oxygenation was achieved.
Table 3. Mean systemic parameters for 6 common FiO2 levels. Values are mean ± standard deviation. * p < 0.05 compared with corresponding 60.5% FIO2 value
3.2 Monte Carlo simulations
The parameters obtained by fitting low SNR spectra (obtained from averaging 16 simulated spectra with added Poisson noise, as described above) at three levels of oxygenation (70, 50 and 35%) are reported in Table 4. Note that the mean and standard deviation displayed in the table were obtained by repeating the simulations 500 times. Analyzing the simulated spectra over a range of averages (1 to 256) and StO2 = 70% yielded coefficient of variations ranging from 0.49 to 0.95%. The mean best-fit estimates for all parameters were similar to the values presented in Table 4, independent of the number of averages. As expected, the uncertainty in each parameter, as reflected by the standard deviation across the 500 trials, was reduced by increasing the number of averages (Fig. 2). This figure indicates that a coefficient of variation (COV) of the order of 5% or less is reached by 64 averages, which represents an acquisition time of the order of 20 s.
Fig. 2. Precision (COV) of the HS-NIRS fitting parameters as a function of SNR (number of averaged spectra)
Table 4. Best-fit estimates from the Monte Carlo simulations
3.3 Optical properties
Table 5 presents the baseline μs′ values measured by TR NIRS and HS NIRS at the initial FiO2 condition (60%). The TR-NIRS values were extracted directly as a fitting parameter, while the broadband values were obtained inputting the fitting parameters β and α, in Eq. 2. For both NIRS methods, μs′ was only calculated at baseline, and the values from the two systems were found to be significantly different across the 4 wavelengths.
Table 5. Baseline reduced scattering coefficients measured by TR and HS NIRS
Figure 3 shows the μa values as a function of wavelength at three of the six oxygenation states spanning the range of conditions (60%, 31% and 20% FiO2). The μa spectra from HS NIRS were generated from Eq. (1) using the measured concentrations of Hb, HbO2 and the water fraction. Overlaid on each spectrum are the μa values from TR NIRS determined at the four wavelengths. The level of agreement between the two methods appeared to remain stable as the oxygenation level decreased. There was a significant difference between the two techniques determined by the repeated measures ANOVA for the 670-nm wavelength only (F1,6 = 24.43, p < 0.05 for 670 nm, F1,6 = 4.62, p > 0.05 for 760 nm, F1,6 = 3.78, p > 0.05 for 804 nm and F1,6 = 0.02, p > 0.05 for 830 nm).
Fig. 3. Absorption coefficient measured by HS NIRS (blue line, grey shadowing is the standard deviation) and TR NIRS (red points, error bars are the standard deviation) across three levels of FiO2: (A) 60.5%, (B) 31% and (C) 20%.
3.4 Cerebral oxygen saturation
Figure 4 shows the concentrations of oxy and deoxy-hemoglobin measured by the two NIRS techniques across six levels of oxygenation. Repeated measures ANOVA revealed no significant difference in HbO2 concentrations between the two NIRS techniques (F1.6 = 1.68, p > 0.05), but there was a significant difference between Hb concentrations across FiO2 values (F1.6 = 39.98, p < 0.05).
Measured cerebral oxygen saturation values from HS and TR NIRS are displayed in Fig. 5 as a function of FiO2. Under the initial FiO2 condition, the average StO2 measures were 76.8 ± 4.0% and 64.6 ± 2.0% from HS and TR NIRS, respectively. Both NIRS techniques exhibited the expected reduction in StO2 with decreasing FiO2. Repeated measures ANOVA revealed that when compared across levels of FiO2, there was a significant effect on the measured StO2 depending on the NIRS technique used (F1,6 = 58.14, p < 0.05). The between-subject COV as determined by a linear mixed effects model across the six common FiO2 conditions was 11.7% for HS NIRS and 12.6% for TR NIRS.
Fig. 5. Plot of the measures of cerebral StO2 from the two NIRS techniques as a function of the FiO2.
The correlation between StO2 measurements from TR and HS NIRS is shown in Fig. 6. The coloured lines show the results of the regression analysis for each animal. The solid black line depicts the average regression line (slope = 1.02 ± 0.28, intercept = 11.78 ± 18.4% and R2 = 0.977 ± 0.035). The average regression slope was found to be significantly different from a slope of zero (p <0.05) but not from the line of identity. In addition, the intercept was significantly different from zero, suggesting a bias between the measurements which was confirmed by the Bland-Altmand analysis. The bias was not dependent on the level of oxygenation, with mean differences and limits of agreement of −12.73 ± 10.4% between the two methods (Fig. 7).
Fig. 6. Regression analysis of StO2 values from HS NIRS and TR NIRS. Analysis was performed for each piglet individually. Average regression results across all piglets is given by the solid black line (slope = 1.02 ± 0.28, intercept = 11.8% ± 18.4% and R2 = 0.977 ± 0.035)
Fig. 7. Bland Altman plot showing the differences in StO2 from the two NIRS techniques. Dashed lines indicate the limits of agreement and the solid line is the mean difference between the techniques and the dotted line is zero.
4. Discussion
This study evaluated the ability of a HS-NIRS technique based on derivative spectroscopy to measure StO2. Acquiring spectral measurements instead of attenuation data at a few discrete wavelengths provides a means of quantifying the main tissue chromophores by modelling the effects of both light absorption and scatter. This approach has shown promise in clinical applications [14], highlighting the need for validation across a range of StO2 values, which was the primary aim of the current study. Experiments were conducted using a piglet model to mimic the small absorption effects of scalp and skull in preterm infants. For validation, StO2 was measured independently by TR NIRS and the comparison was performed over a range of StO2 values by altering FiO2 from 20% to 60%.
To assess the fitting algorithm independent of any potential experimental errors, simulated spectra were generated at three different StO2 values (35%, 50% and 70%) with noise characteristics comparable to real data. The results of the simulations demonstrated that the algorithm provided accurate estimates of all five fitting parameters since there was no bias between the mean estimates and their corresponding true input values (Table 4 versus Table 2). In terms of precision, the COV across the five parameters ranged from 1.6% to 6.7%, with the scattering parameter β having the greatest COV across the three oxygenation levels. Interestingly, the COV of StO2 was smaller than the corresponding COVs for Hb and HbO2, indicating that the errors in the two fitting parameters are not uncorrelated. That is, if noise causes an underestimation of Hb, then likely HbO2 will be underestimated as well, resulting in less error in the derived StO2 value. Repeating the simulations across a range of averages (1 to 256) demonstrated that the fitting routine is robust. When comparing the StO2 measurements, the COV was below 1% for all levels of SNR with values of 0.95%, 0.59% and 0.49% for 1, 16 and 256 averages respectively. To put these levels in context, HS-NIRS data from the initial FiO2 condition in the experimental data were divided into segments of 16 spectra that were fit individually. The average COV from the experimental data was 0.56 ± 0.16%, which was in good agreement with the Monte Carlo simulations.
In terms of the experimental results, the main finding was a strong linear correlation between StO2 estimates from the two optical techniques across a range of FiO2 values. The average coefficient of determination obtained by comparing StO2 values from HS and TR NIRS was 0.95 ± 0.075, and the average regression slope was 1.02 ± 0.28. A high coefficient of determination was found in all experiments with values ranging from 0.900 to 0.995. These results demonstrate that the two techniques are comparable in terms of their ability to track reductions in StO2 caused by increasing levels of hypoxia. This is important, since the main use of the HS-NIRS technique is to determine StO2, which can be used solely or in combination of cerebral perfusion to compute cerebral oxidative metabolism [23]. Despite this good agreement, discrepancies between the two methods were found in terms of mean baseline StO2 (76.8 ± 4.0% and 64.6 ± 2.0% for HS and TR NIRS, respectively) and μs′ values (Table 5). The average baseline TR-NIRS StO2 value was in good agreement with previous studies involving neonatal piglets [15,28,29]. In particular, Verdecchia et al. reported a baseline saturation of 69% using a three channel TR-NIRS system similar to the unit used in the current study [17]. Likewise, lower μs′ values have been reported in previous HS-NIRS studies [11,30].
These discrepancies are not related to the fitting routine as demonstrated by the results from the computer simulations, which indicated that all of the fitting parameters could be estimated accurately. Since derivative spectroscopy relies on characterizing spectra features, its accuracy can be negatively impacted by factors that cause unanticipated spectral distortions. One potential issue is incorrectly measuring the reference spectrum, which is needed to account for spectral effects of the instrumentation, including the shape of the emission spectrum and wavelength-variations in detection efficiency of the spectrometer. However, we do not believe the distortions in the measured reference spectrum would likely explain the discrepancies found in the current study considering that the lower μs′ values are in agreement with previous HS-NIRS studies involving different spectrometers [11].
An alternative explanation is light absorption by tissue that is not accounted for by modelling μa(λ) as the sum of Hb, HO2 and water. The presence of background absorption was reported by Ijichi et al. [28] who observed substantial absorption at all measured wavelengths (761, 795, and 835 nm) after performing a blood exchange with a perfluorotributylamine emulsion in newborn piglets. That is, the background measurement of the piglet head was different from that expected assuming only absorption from 85% water. Ijichi found that including this background absorption in the definition of μa (λ) reduced the estimates of Hb and HbO2 concentrations. Applying the same correction to the TR-NIRS data from the current study lowered the mean HbO2 concentration from 39 to 23 µM and the Hb concentration from 19 to 12 µM. The adjusted Hb concentration is more aligned with previous studies [4] and in good agreement with the estimate from HS NIRS (13 ± 2.7 µM). However, this correction led to a greater discordance between HbO2 values from HS and TR NIRS. Considering that any unaccounted-for tissue absorption would have greater effects on the first derivative spectrum rather than the second, errors in HbO2 and μs′ would be more likely than errors in Hb. The possibility of such distortions due to background absorption requires verification, such as repeating the blood exchange experiments with HS-NIRS measurements. Another approach would be to use tissue-mimicking optical phantoms in which the oxygenation levels of added hemoglobin can be controlled [31]. However, applying derivative spectroscopy requires prior knowledge of the absorption spectrum of the lipid emulsion (e.g., Intralipid). Since it is not the same as for water, this requires characterizing the scatting and absorption spectra at a wavelength resolution sufficient for derivative spectroscopy [32,33].
An additional contributing factor to the discrepancy between StO2 estimates from the two techniques could be signal contamination from the extracerebral layers, which would likely affect the techniques differently considering HS NIRS is more sensitive to superficial layers. This potential error is generally ignored with piglets since the thickness of the extracerebral layer is in the range of 2 to 3 mm [23], which is small compared to a source-detector separation of 30 mm. Previous studies have reported that cerebral blood flow estimates derived from HS NIRS were in good agreement with measurements from other perfusion techniques [5] [23], suggesting that this assumption is reasonable. However, these blood flow measurements are based on a dynamic contrast-enhanced technique, which may have a different depth sensitivity compared to the derivative approach used to determine StO2. Therefore, this potential source of error cannot be ruled out. Enhancing depth sensitivity could be achieved by collecting spectra at multiple source-detector distances, analogous to spatially resolved NIRS. This approach would also be advantageous for studies involving term infants considering the increase in head size relative to preterm infants.
There are a number of potential limitations with this study. First, the fitting approaches for both techniques only included light-scattering terms in the analysis of baseline data and not subsequent FiO2 steps. This approach was used to reduce the variability in the estimates considering hypoxia only affects blood oxygenation and not the tissue scattering properties [34] nor the water fraction [35]. Second, this study did not include repeat measurements to assess reproducibility, as has been performed in clinical studies using commercial NIRS devices [36]. Typically, these studies involve repetitively placing the NIRS probes on the patient’s head to acquire serial measurements. Practically, this would have been challenging to perform in the current experiments given the small size of the piglet head, which makes replicating probe placement challenging. However, given the encouraging results of this validation study, “place and replace” acquisitions in clinical studies to assess precision is a logical next step. A final potential limitation is the assumption of a homogenous medium.
In summary, the work presented in this study demonstrated the ability of HS NIRS to measure StO2 over a wide range of oxygenation states in a piglet model of the neonatal head. The measured changes in StO2 were in strong agreement with those obtained with TR NIRS, which is considered the gold standard for measuring StO2 given its ability to separate absorption and scattering effects. Although a strong correlation between StO2 values from both techniques was found in each experiment, the study revealed significant discrepancies in μs′ and StO2 values from HS NIRS compared to estimates from TR NIRS. It was speculated that these differences might have been caused by background absorption not accounted for in the spectral modelling. Despite these differences, HS NIRS provides some unique advantages in terms of neonatal brain monitoring given the simplicity of the instrumentation, low cost, and its ability to quantify the concentrations of multiple chromophores.
Funding
Canadian Institutes of Health Research (140171); Natural Sciences and Engineering Research Council of Canada (478470-2015).
Acknowledgments
The authors would like to thank Jennifer Hadway and Lise Desjardins for help in conducting animal experiments and Lynn Keenliside for technical support.
Disclosures
The authors declare that there are no conflicts of interest related to this article.
References
1. M. Kenosi, G. Naulaers, C. A. Ryan, and E. M. Dempsey, “Current research suggests that the future looks brighter for cerebral oxygenation monitoring in preterm infants,” Acta Paediatr. 104(3), 225–231 (2015). [CrossRef]
2. S. Hyttel-Sorensen, A. Pellicer, T. Alderliesten, T. Austin, F. van Bel, M. Benders, O. Claris, E. Dempsey, A. R. Franz, M. Fumagalli, C. Gluud, B. Grevstad, C. Hagmann, P. Lemmers, W. van Oeveren, G. Pichler, A. M. Plomgaard, J. Riera, L. Sanchez, P. Winkel, M. Wolf, and G. Greisen, “Cerebral near infrared spectroscopy oximetry in extremely preterm infants: phase II randomised clinical trial,” The BMJ 350(jan05 2), g7635 (2015). [CrossRef]
3. S. J. Matcher, M. Cope, and D. T. Delpy, “Use of the water absorption spectrum to quantify tissue chromophore concentration changes in near-infrared spectroscopy,” Phys. Med. Biol. 39(1), 177–196 (1994). [CrossRef]
4. C. E. Cooper, C. E. Elwell, J. H. Meek, S. J. Matcher, J. S. Wyatt, M. Cope, and D. T. Delpy, “The Noninvasive Measurement of Absolute Cerebral Deoxyhemoglobin Concentration and Mean Optical Path Length in the Neonatal Brain by Second Derivative Near Infrared Spectroscopy,” Pediatr. Res. 39(1), 32–38 (1996). [CrossRef]
5. D. W. Brown, P. A. Picot, J. Gharavi Naeini, R. Springett, D. T. Delpy, and T.-Y. Lee, “Quantitative Near Infrared Spectroscopy Measurement of Cerebral Hemodynamics in Newborn Piglets,” Pediatr. Res. 51(5), 564–570 (2002). [CrossRef]
6. K. M. Tichauer, J. A. Hadway, T.-Y. Lee, and K. S. Lawrence, “Measurement of Cerebral Oxidative Metabolism with Near-Infrared Spectroscopy: A Validation Study,” J. Cereb. Blood Flow Metab. 26(5), 722 (2006). [CrossRef]
7. R. Arora, M. Ridha, D. S. C. Lee, J. Elliott, H. C. Rosenberg, M. Diop, T.-Y. Lee, and K. St. Lawrence, “Preservation of the metabolic rate of oxygen in preterm infants during indomethacin therapy for closure of the ductus arteriosus,” Pediatr. Res. 73(6), 713–718 (2013). [CrossRef]
8. G. Bale, C. E. Elwell, and I. Tachtsidis, “From Jöbsis to the present day: a review of clinical near-infrared spectroscopy measurements of cerebral cytochrome-c-oxidase,” J. Biomed. Opt. 21(9), 091307 (2016). [CrossRef]
9. G. Bale, S. Mitra, I. de Roever, M. Sokolska, D. Price, A. Bainbridge, R. Gunny, C. Uria-Avellanal, G. S. Kendall, J. Meek, N. J. Robertson, and I. Tachtsidis, “Oxygen dependency of mitochondrial metabolism indicates outcome of newborn brain injury,” J. Cereb. Blood Flow Metab. 0271678X1877792 (2018).
10. O. Pucci, V. Toronov, and K. S. Lawrence, “Measurement of the optical properties of a two-layer model of the human head using broadband near-infrared spectroscopy,” Appl. Opt. 49(32), 6324–6332 (2010). [CrossRef]
11. H. Z. Yeganeh, V. Toronov, J. T. Elliott, M. Diop, T.-Y. Lee, and K. St. Lawrence, “Broadband continuous-wave technique to measure baseline values and changes in the tissue chromophore concentrations,” Biomed. Opt. Express 3(11), 2761 (2012). [CrossRef]
12. M. Diop, E. Wright, V. Toronov, T.-Y. Lee, and K. S. Lawrence, “Improved light collection and wavelet de-noising enable quantification of cerebral blood flow and oxygen metabolism by a low-cost, off-the-shelf spectrometer,” J. Biomed. Opt. 19(5), 057007 (2014). [CrossRef]
13. M. Diop, J. Kishimoto, V. Toronov, D. S. C. Lee, and K. St. Lawrence, “Development of a combined broadband near-infrared and diffusion correlation system for monitoring cerebral blood flow and oxidative metabolism in preterm infants,” Biomed. Opt. Express 6(10), 3907 (2015). [CrossRef]
14. P. J. McLachlan, J. Kishimoto, M. Diop, D. Milej, D. S. C. Lee, S. de Ribaupierre, and K. St Lawrence, “Investigating the effects of cerebrospinal fluid removal on cerebral blood flow and oxidative metabolism in infants with post-hemorrhagic ventricular dilatation,” Pediatr. Res. 82(4), 634–641 (2017). [CrossRef]
15. A. Rajaram, G. Bale, M. Kewin, L. B. Morrison, I. Tachtsidis, K. St. Lawrence, and M. Diop, “Simultaneous monitoring of cerebral perfusion and cytochrome c oxidase by combining broadband near-infrared spectroscopy and diffuse correlation spectroscopy,” Biomed. Opt. Express 9(6), 2588 (2018). [CrossRef]
16. K. M. Tichauer, J. T. Elliott, J. A. Hadway, D. S. Lee, T.-Y. Lee, and K. St. Lawrence, “Using near-infrared spectroscopy to measure cerebral metabolic rate of oxygen under multiple levels of arterial oxygenation in piglets,” J. Appl. Physiol. 109(3), 878–885 (2010). [CrossRef]
17. K. Verdecchia, M. Diop, T.-Y. Lee, and K. S. Lawrence, “Quantifying the cerebral metabolic rate of oxygen by combining diffuse correlation spectroscopy and time-resolved near-infrared spectroscopy,” J. Biomed. Opt. 18(2), 027007 (2013). [CrossRef]
18. D. Milej, A. Abdalmalak, P. McLachlan, M. Diop, A. Liebert, and Keith St. Lawrence, “Subtraction-based approach for enhancing the depth sensitivity of time-resolved NIRS,” Biomed. Opt. Express 7(11), 4514–4526 (2016). [CrossRef]
19. D. Milej, A. Abdalmalak, L. Desjardins, H. Ahmed, T.-Y. Lee, M. Diop, and K. St. Lawrence, “Quantification of blood-brain barrier permeability by dynamic contrast-enhanced NIRS,” Sci. Rep. 7(1), 1702 (2017). [CrossRef]
20. A. Abdalmalak, D. Milej, L. Norton, D. B. Debicki, T. Gofton, M. Diop, A. M. Owen, and K. St. Lawrence, “Single-session communication with a locked-in patient by functional near-infrared spectroscopy,” Neurophotonics 4(04), 1 (2017). [CrossRef]
21. M. Diop, K. M. Tichauer, J. T. Elliott, M. Migueis, T.-Y. Lee, and K. St. Lawrence, “Comparison of time-resolved and continuous-wave near-infrared techniques for measuring cerebral blood flow in piglets,” J. Biomed. Opt. 15(5), 057004 (2010). [CrossRef]
22. A. Kienle and M. S. Patterson, “Improved solutions of the steady-state and the time-resolved diffusion equations for reflectance from a semi-infinite turbid medium,” J. Opt. Soc. Am. A 14(1), 246–254 (1997). [CrossRef]
23. M. Diop, J. T. Elliott, K. M. Tichauer, T.-Y. Lee, and K. St. Lawrence, “A broadband continuous-wave multichannel near-infrared system for measuring regional cerebral blood flow and oxygen consumption in newborn piglets,” Rev. Sci. Instrum. 80(5), 054302 (2009). [CrossRef]
24. C. Lau, O. Šćepanović, J. Mirkovic, S. McGee, C.-C. Yu, S. Fulghum, M. Wallace, J. Tunnell, K. Bechtel, and M. Feld, “Re-evaluation of model-based light-scattering spectroscopy for tissue spectroscopy,” J. Biomed. Opt. 14(2), 024031 (2009). [CrossRef]
25. M. Diop and K. St. Lawrence, “Deconvolution method for recovering the photon time-of-flight distribution from time-resolved measurements,” Opt. Lett. 37(12), 2358 (2012). [CrossRef]
26. V. Ntziachristos, A. G. Yodh, and B. Chance, “Accuracy limits in the determination of absolute optical properties using time-resolved NIR spectroscopy,” in (1999) 3597, pp. 3597–10.
27. E. Alerstam, S. Andersson-Engels, and T. Svensson, “Improved accuracy in time-resolved diffuse reflectance spectroscopy,” Opt. Express 16(14), 10440 (2008). [CrossRef]
28. S. Ijichi, T. Kusaka, K. Isobe, F. Islam, K. Okubo, H. Okada, M. Namba, K. Kawada, T. Imai, and S. Itoh, “Quantification of cerebral hemoglobin as a function of oxygenation using near-infrared time-resolved spectroscopy in a piglet model of hypoxia,” J. Biomed. Opt. 10(2), 024026 (2005). [CrossRef]
29. D. M. Hueber, M. A. Franceschini, H. Y. Ma, Q. Zhang, J. R. Ballesteros, S. Fantini, D. Wallace, V. Ntziachristos, and B. Chance, “Non-invasive and quantitative near-infrared haemoglobin spectrometry in the piglet brain during hypoxic stress, using a frequency-domain multidistance instrument†,” Phys. Med. Biol. 46(1), 41–62 (2001). [CrossRef]
30. S. L. Jacques, “Optical properties of biological tissues: a review,” Phys. Med. Biol. 58(11), R37–R61 (2013). [CrossRef]
31. S. Kleiser, D. Ostojic, B. Andresen, N. Nasseri, H. Isler, F. Scholkmann, T. Karen, G. Greisen, and M. Wolf, “Comparison of tissue oximeters on a liquid phantom with adjustable optical properties: an extension,” Biomed. Opt. Express 9(1), 86–101 (2018). [CrossRef]
32. P. Taroni, A. Pifferi, E. Salvagnini, L. Spinelli, A. Torricelli, and R. Cubeddu, “Seven-wavelength time-resolved optical mammography extending beyond 1000 nm for breast collagen quantification,” Opt. Express 17(18), 15932 (2009). [CrossRef]
33. S. Grabtchak, “Experimental spectro-angular mapping of light distribution in turbid media,” J. Biomed. Opt. 17(6), 067007 (2012). [CrossRef]
34. G. Zhang, A. Katz, R. R. Alfano, A. D. Kofinas, D. A. Kofinas, P. G. Stubblefield, W. Rosenfeld, D. Beyer, D. Maulik, and M. R. Stankovic, “Brain perfusion monitoring with frequency-domain and continuous-wave near-infrared spectroscopy: a cross-correlation study in newborn piglets,” Phys. Med. Biol. 45(11), 3143–3158 (2000). [CrossRef]
35. A. Demel, M. Wolf, C. F. Poets, and A. R. Franz, “Effect of different assumptions for brain water content on absolute measures of cerebral oxygenation determined by frequency-domain near-infrared spectroscopy in preterm infants: an observational study,” BMC Pediatr. 14(1), 206 (2014). [CrossRef]
36. C. Jenny, M. Biallas, I. Trajkovic, J.-C. Fauchère, H. U. Bucher, and M. Wolf, “Reproducibility of cerebral tissue oxygen saturation measurements by near-infrared spectroscopy in newborn infants,” J. Biomed. Opt. 16(9), 097004 (2011). [CrossRef]
