Assessing the Sensitivity of Multi-Distance Hyperspectral NIRS to Changes in the Oxidation State of Cytochrome C Oxidase in the Brain

Near-infrared spectroscopy (NIRS) measurements of tissue oxygen saturation (StO2) are frequently used during vascular and cardiac surgeries as a non-invasive means of assessing brain health; however, signal contamination from extracerebral tissues remains a concern. As an alternative, hyperspectral (hs)NIRS can be used to measure changes in the oxidation state of cytochrome c oxidase (ΔoxCCO), which provides greater sensitivity to the brain given its higher mitochondrial concentration versus the scalp. The purpose of this study was to evaluate the depth sensitivity of the oxCCO signal to changes occurring in the brain and extracerebral tissue components. The oxCCO assessment was conducted using multi-distance hsNIRS (source-detector separations = 1 and 3 cm), and metabolic changes were compared to changes in StO2. Ten participants were monitored using an in-house system combining hsNIRS and diffuse correlation spectroscopy (DCS). Data were acquired during carotid compression (CC) to reduce blood flow and hypercapnia to increase flow. Reducing blood flow by CC resulted in a significant decrease in oxCCO measured at rSD = 3 cm but not at 1 cm. In contrast, significant changes in StO2 were found at both distances. Hypercapnia caused significant increases in StO2 and oxCCO at rSD = 3 cm, but not at 1 cm. Extracerebral contamination resulted in elevated StO2 but not oxCCO after hypercapnia, which was significantly reduced by applying regression analysis. This study demonstrated that oxCCO was less sensitive to extracerebral signals than StO2.


Introduction
Various procedures employed during cardiac and vascular surgery, such as cardiopulmonary bypass or arterial clamping, place the patient at risk of brain injury, with an incidence of postoperative stroke between 0.8% and 5.2% [1][2][3][4] and cognitive decline between 14.1% and 50% [5,6]. In an effort to reduce the risk of neurological complications, brain monitoring has become an essential component of intraoperative management. Several techniques have been evaluated, including electroencephalography (EEG) [7], somatosensory evoked potential (SEP) [8], transcranial Doppler (TCD) [9], and cerebral tissue oxygen saturation (StO 2 ) by near-infrared spectroscopy (NIRS) [10]. A disadvantage of EEG and SEP is that the signals only indirectly reflect cerebral blood flow (CBF) [11]. While a decrease in amplitude in EEG or SEP can indicate reduced CBF, not all EEG and SEP changes are associated with ischemic injury, and stroke can occur even in the absence of changes [12]. Unnecessary shunt placement during carotid endarterectomies is associated with TCD

Experimental Protocol
All experiments were approved by the Western University Health Sciences Research Ethics Board, which adheres to the guidelines of the Tri-Council Policy Statement for research involving humans. Written informed consent was obtained from each participant before the experiment. Volunteers were excluded based on a neurological or psychiatric disorder diagnosis or a history of vascular disease. All participants completed both the carotid compression and hypercapnia experiments.
Participants were studied in the supine position. Optical probes were attached to the right side of their forehead via a custom-designed probe holder secured by a Velcro headband. One detection fiber bundle was placed at r SD = 1 cm, and three detection fiber bundles, which collected both NIRS and DCS signals, were placed at r SD = 3 cm from the NIRS source and 2 cm from the DCS source (Figure 1), respectively. fiber (Φ = 600 μm, NA = 0.39; Thorlabs, Newton, NJ, USA) and transmitted through a shutter to a spectrometer (AvaSpec-ULS2048XL, λBandwith = 666-1025 nm, λResolution = 0.18 nm; Avantes, The Netherlands). At rSD = 3 cm, reflected light was collected by three fiber bundles (Ø = 2 mm, Φ = 30 μm, NA = 0.55; Loptek, Germany) that were linearly aligned at the entrance of a second spectrometer (iDus 420, λBandwith = 548-1081 nm, λResolution = 0.52 nm; Andor, Oxford Instruments, Toronto, ON, Canada). Reflectance spectra were acquired at both distances simultaneously.

Experimental Protocol
All experiments were approved by the Western University Health Sciences Research Ethics Board, which adheres to the guidelines of the Tri-Council Policy Statement for research involving humans. Written informed consent was obtained from each participant before the experiment. Volunteers were excluded based on a neurological or psychiatric disorder diagnosis or a history of vascular disease. All participants completed both the carotid compression and hypercapnia experiments.
Participants were studied in the supine position. Optical probes were attached to the right side of their forehead via a custom-designed probe holder secured by a Velcro headband. One detection fiber bundle was placed at rSD = 1 cm, and three detection fiber bundles, which collected both NIRS and DCS signals, were placed at rSD = 3 cm from the NIRS source and 2 cm from the DCS source (Figure 1), respectively. Continuous arterial blood pressure was measured by finger photoplethysmography (Finometer, Finapres Medical Systems, Enschede, Netherlands), which was calibrated Continuous arterial blood pressure was measured by finger photoplethysmography (Finometer, Finapres Medical Systems, Enschede, Netherlands), which was calibrated against three manual measurements for the sphygmomanometric brachial artery. Arterial blood pressure was used to calculate mean arterial pressure (MAP).

Carotid Compressions (CC)
The experimental protocol for CC consisted of a 1-min baseline period followed by three digital compressions of the right (i.e., ipsilateral to the position of the probes) common carotid artery at the level of 1 cm superior to the clavicle ( Figure 2) [39]. Each compression lasted for 15 s, followed by a 30-s recovery period. The procedure was then performed on the left common carotid artery (i.e., contralateral to the position of the probes). Finally, compression was repeated on the right common carotid artery for a single 30-s period, followed by 1.5 min of recovery.

Carotid Compressions (CC)
The experimental protocol for CC consisted of a 1-min baseline period followed by three digital compressions of the right (i.e., ipsilateral to the position of the probes) common carotid artery at the level of 1 cm superior to the clavicle ( Figure 2) [39]. Each compression lasted for 15 s, followed by a 30-s recovery period. The procedure was then performed on the left common carotid artery (i.e., contralateral to the position of the probes). Finally, compression was repeated on the right common carotid artery for a single 30-s period, followed by 1.5 min of recovery.
The hsNIRS/DCS system enables hsNIRS and DCS data to be collected sequentially using a multiplexing shutter system; however, due to the rapid response to CC, hsNIRS and DCS data were collected separately during these experiments.

Hypercapnia
Subjects were required to wear a facemask connected to a computer-controlled gas delivery circuit (RespirAct, Thornhill Research Inc, Toronto, ON, Canada). The experimental protocol consisted of one 4-min period of hypercapnia in which end-tidal carbon dioxide pressure (PETCO2) was increased by 10 mmHg above each subject's normocapnic PETCO2 value, as determined by the gas delivery circuit. The hypercapnic period started two minutes after baseline recordings and was followed by three minutes of normocapnia. Hyperspectral NIRS and DCS data were recorded sequentially during the experiment.

Hyperspectral NIRS
At the beginning of each experiment [40], a reference spectrum (referenceλ) and a dark count spectrum (darkλ) were acquired for each spectrometer (i.e., one at rSD = 1 cm and the other at rSD = 3 cm). Spectra (dataλ) collected during the baseline period before either CC or hypercapnia were converted into baseline reflectance spectra using the following: The first and second derivatives of R(λ) were fitted with the solution to the diffusion approximation for a semi-infinite homogeneous medium [41] to generate estimates of the tissue water fraction, HbO2 and Hb concentrations, and two scattering parameters (wavelength-dependent power and the reduced scattering coefficient (µs′) at 800 nm) [36]. Fitting was performed using a constrained minimization algorithm based on the MATLAB function fminsearchbnd (2016b, MathWorks, USA). The HbO2 and Hb concentrations estimates were used to calculate baseline tissue oxygen saturation at rSD = 1 and 3 cm. The hsNIRS/DCS system enables hsNIRS and DCS data to be collected sequentially using a multiplexing shutter system; however, due to the rapid response to CC, hsNIRS and DCS data were collected separately during these experiments.

Hypercapnia
Subjects were required to wear a facemask connected to a computer-controlled gas delivery circuit (RespirAct, Thornhill Research Inc, Toronto, ON, Canada). The experimental protocol consisted of one 4-min period of hypercapnia in which end-tidal carbon dioxide pressure (P ET CO 2 ) was increased by 10 mmHg above each subject's normocapnic P ET CO 2 value, as determined by the gas delivery circuit. The hypercapnic period started two minutes after baseline recordings and was followed by three minutes of normocapnia. Hyperspectral NIRS and DCS data were recorded sequentially during the experiment.

Hyperspectral NIRS
At the beginning of each experiment [40], a reference spectrum (reference λ ) and a dark count spectrum (dark λ ) were acquired for each spectrometer (i.e., one at r SD = 1 cm and the other at r SD = 3 cm). Spectra (data λ ) collected during the baseline period before either CC or hypercapnia were converted into baseline reflectance spectra using the following: The first and second derivatives of R(λ) were fitted with the solution to the diffusion approximation for a semi-infinite homogeneous medium [41] to generate estimates of the tissue water fraction, HbO 2 and Hb concentrations, and two scattering parameters (wavelength-dependent power and the reduced scattering coefficient (µ s ) at 800 nm) [36]. Fitting was performed using a constrained minimization algorithm based on the MATLAB function fminsearchbnd (2016b, MathWorks, USA). The HbO 2 and Hb concentrations estimates were used to calculate baseline tissue oxygen saturation at r SD = 1 and 3 cm.
Changes in Hb, HbO 2 , and oxCCO concentrations relative to their baseline values were estimated using the modified Beer-Lambert law based on the UCLn algorithm [18]. The analysis was performed separately for spectra acquired at r SD = 1 and 3 cm. Changes in Hb and HbO 2 concentrations were determined from attenuation changes measured between λ = 680 and 850 nm [42]. Likewise, changes in oxCCO concentration were determined from attenuation changes between λ = 770 and 906 nm. For this analysis, the differential pathlength for each subject was calculated by fitting the second derivative of average baseline R(λ) to the second derivative of the water absorption spectrum [43] and correcting for the wavelength dependence of the pathlength [44]. StO 2 at each time point was determined by combining the relative changes derived from the UCLn algorithm with the absolute baseline value obtained by derivative spectroscopy. The StO 2 time courses were smoothed with a zero-phase digital filter (filtfilt, MATLAB, 2016b, MathWorks, Natick, MA, USA).

DCS
Using the Siegert relation, normalized intensity autocorrelations functions were converted to electric field autocorrelation data [45]. Each autocorrelation function was subsequently fit with the diffusion approximation solution for a semi-infinite homogenous medium. The fitting incorporated assumed values of the optical coefficients µ a = 0.1 cm −1 and µ s = 10 cm −1 [46] and the coherence factor (β) determined from the average initial value of the baseline g 2 curves. The fitting procedure was performed across all correlation times from 1 µs to 1 ms and yielded a best-fit estimate of the blood flow index (BFi) based on modelling tissue perfusion as pseudo-Brownian motion [47]. The resulting BFi time courses were smoothed with the same filter applied to the hsNIRS data (i.e., zero-phase digital filtering; filtfilt, MATLAB, 2016b, MathWorks, USA).

Cerebrovascular Reactivity
To determine the response time of ∆BFi, ∆StO 2 , and ∆oxCCO to 30-s CC, the time courses of ∆BFi, ∆StO 2 , and ∆oxCCO were modelled as the convolution of a step function representing carotid compression (denoted CC(t) and scaled to a maximum value of one) and a hemodynamic response function (HRF) [33,48]: where ∆S(t) is the signal change, ssCVR is the steady-state value of cerebrovascular reactivity (CVR) and * denotes the convolution operator. The HRF is given by: where τ is the time constant defining the dynamic component of CVR, t 0 is the time delay between the start of CC(t) and the initial decline of ∆S(t), and N is the area under ∞ 0 e −t/τ dt. Best-fit estimates of τ, t 0 , and ssCVR were obtained by numerical optimization (fminsearchbnd, MATLAB, Mathworks Inc., USA). The fitting was performed over a time window spanning the beginning of CC to the nadir of ∆S(t). For ∆StO 2 and ∆oxCCO, the analysis was performed for time courses recorded at r SD = 3 cm.

Statistical Analysis
All data are presented as mean ± standard deviation unless otherwise noted. Statistical analyses were conducted in IBM SPSS. Statistical significance was defined as p < 0.05. Multivariate analyses of variance (ANOVA) were used to compare ∆StO 2 and ∆oxCCO at the two r SD (1 and 3 cm) during the two compression durations (15 and 30 s). Independent-samples t-tests were used to evaluate ∆StO 2 and ∆oxCCO at the two r SD (1 and 3 cm) and ∆StO 2.3cm and ∆oxCCO 3cm versus ∆StO 2 , Reg and ∆oxCCO Reg . Paired-samples t-tests were used to evaluate ∆StO 2,1cm , ∆oxCCO 1cm , ∆BFi, and change in MAP versus the baseline. A repeated measures ANOVA was conducted on the 15-s ipsilateral CC data to determine the precision of ∆oxCCO and ∆StO 2 . Precision was defined by the coefficient of variation (CoV) for the within-subject variance.

Results
Data were acquired from 10 participants (4 females, 6 males, 24-34 years, mean = 29 ± 5 years). A total of 67 digital common carotid artery compressions were performed (30 15-s right CCs, 28 15-s left CCs, and 10 30-s right CCs). Data from one participant were excluded from the contralateral 15-s CC analysis as the participant experienced mild syncopal symptoms during the contralateral 15-s CC. The same 10 participants also underwent the hypercapnia protocol. Figure 3 presents time courses of average changes in BFi, StO 2 , and oxCCO in response to ipsilateral CC across subjects during the two compression durations. Data for ∆StO 2 and ∆oxCCO are presented for both source-detector separations (r SD = 1 and 3 cm). Decreases in ∆BFi, ∆StO 2 , and ∆oxCCO were observed at both source-detector distances during ipsilateral 15 s and 30 s CC. Change in each parameter in response to CC was characterized by taking the average of 5 s around the maximum reduction (Table 1).

Carotid Compressions (CC)
the analysis was performed for time courses recorded at rSD = 3 cm.

Statistical Analysis
All data are presented as mean ± standard deviation unless otherwise noted. Statistical analyses were conducted in IBM SPSS. Statistical significance was defined as p < 0.05. Multivariate analyses of variance (ANOVA) were used to compare ΔStO2 and ΔoxCCO at the two rSD (1 and 3 cm) during the two compression durations (15 and 30 s). Independentsamples t-tests were used to evaluate ΔStO2 and ΔoxCCO at the two rSD (1 and 3 cm) and ΔStO2.3cm and ΔoxCCO3cm versus ΔStO2,Reg and ΔoxCCOReg. Paired-samples t-tests were used to evaluate ΔStO2,1cm, ΔoxCCO1cm, ΔBFi, and change in MAP versus the baseline. A repeated measures ANOVA was conducted on the 15-s ipsilateral CC data to determine the precision of ∆oxCCO and ∆StO2. Precision was defined by the coefficient of variation (CoV) for the within-subject variance.
Thirty-second CC resulted in a significant decrease in BFi (−57 ± 14%) and an increase in MAP (4 ± 1 mmHg). Decreases in oxCCO recorded at r SD = 3 cm (Table 1) were significantly larger than the reductions measured at 1 cm (−0.06 ± 0.1 µM). The corresponding Metabolites 2022, 12, 817 7 of 14 decrease in StO 2 at the longer r SD (−4 ± 2.2%) was also significantly larger than the decrease recorded at r SD = 1 cm. The reduction in StO 2 recorded at 1 cm was significantly less than baseline, whereas the reduction in oxCCO at 1 cm did not reach significance.
The 15-s CC response was averaged over the three trials for every subject. The average significant decrease in BFi was 55 ± 8%, and an increase in MAP (4 ± 3 mmHg). Similar to 30-s CC, reductions in ∆oxCCO and ∆StO 2 recorded at r SD = 3 cm were significantly greater than the corresponding reductions measured at 1 cm. StO 2 and oxCCO changes measured for 15-s CC were not statistically different from those obtained for the 30-s CC. From the three 15-s CC trials, the estimated CoV for within-subject variability was 6% and 1% for ∆oxCCO at r SD = 1 and 3 cm, respectively. Similar values were found for the corresponding ∆StO 2 measurements: CoV = 9% and 8% at r SD = 1 and 3 cm, respectively.
Fitting the cerebrovascular reactivity model to the time courses for 30-s CC demonstrated that ∆BFi exhibited the fastest response and ∆StO 2 the slowest, as indicated by the time constant defining HRF(t); i.e., τ = 1.8 ± 1.4 s for ∆BFi, 4.8 ± 3.5 s for ∆oxCCO, and 14.8 ± 8.4 s for ∆StO 2 . The average τ value for ∆StO 2 was significantly different from the corresponding values for ∆BFi and ∆oxCCO, while the values for ∆oxCCO and ∆BFi were not significantly different from each other. Despite the lack of significant difference between the response time between ∆BFi and ∆oxCCO, an average temporal delay of 4.7 ± 7.3 s was found between the nadirs.
Following completion of carotid compression, a brief 5-s hyperemic response was observed, which was characterized by a BFi increase of 32 ± 20% after 30-s CC and 28 ± 26% after 15-s CC; however, there was no significant change in ∆StO 2 and ∆oxCCO. Figure 4 presents average time courses of ∆StO 2 and ∆oxCCO measured at r SD = 3 cm in response to 30-s CC after applying regression analysis. In both cases, regression reduced the magnitude of the response to CC. Both reductions remained significantly different from baseline after regression. The maximum decrease was 2.4 ± 1.9% for ∆StO 2 (Figure 4a) and 0.21 ± 0.24 µM for ∆oxCCO (Figure 4b). Regression also significantly reduced the time constant for regressed ∆StO 2 (τ = 6.6 ± 5.6 s), and the average τ value for regressed ∆StO 2 was not significantly different from the τ values for ∆BFi and ∆oxCCO.     1 and 1.5 min). In both cases, the time course measured at r SD = 1 cm was used as the regressor. Time courses were averaged across subjects, and shading surrounding each line represents the standard deviation. Figure 5 presents the correlation of ∆oxCCO to ∆BFi during CC. A strong non-linear relationship can be observed for ∆oxCCO recorded at r SD = 3 cm, with all ∆oxCCO values for BFi ≥ 21% significantly different from zero. In contrast, ∆oxCCO recorded at r SD at 1 cm was relatively unresponsive to ∆BFi, with no values significantly different from zero. each line represents the standard deviation. Figure 5 presents the correlation of ΔoxCCO to ∆BFi during CC. A strong non-linear relationship can be observed for ΔoxCCO recorded at rSD = 3 cm, with all ΔoxCCO values for BFi ≥ 21% significantly different from zero. In contrast, ΔoxCCO recorded at rSD at 1 cm was relatively unresponsive to ∆BFi, with no values significantly different from zero. Small decreases in ΔStO2 and ΔoxCCO were observed on the contralateral hemisphere in response to CC of 15 s ( Figure 6); however, these changes did not reach significance. In contrast, a significant increase in BFi (14 ± 14%) was found. Small decreases in ∆StO 2 and ∆oxCCO were observed on the contralateral hemisphere in response to CC of 15 s ( Figure 6); however, these changes did not reach significance. In contrast, a significant increase in BFi (14 ± 14%) was found.  Figure 7 presents average time courses of changes in BFi, StO2, and oxCCO in response to 4 min of hypercapnia (PETCO2 increase = 10 ± 2 mmHg). Hypercapnic responses were calculated as the relative difference between the average signal from 4 to 6 min (i.e., 2nd half of the hypercapnic period) and the average of the first minute of baseline. Average ΔoxCCO and ΔStO2 recorded at both source-detectors distances are provided in Table  2. ΔoxCCO and ΔStO2 measured at rSD = 3 cm were significantly larger than the responses measured at rSD = 1 cm. The ΔoxCCO and ΔStO2 responses at 1 cm were significantly delayed (33 ± 24 s and 10 ± 6 s, respectively) compared to the corresponding responses at 3 cm. Hypercapnia resulted in a significant increase in BFi (31 ± 48%) and MAP (4 ± 1 mmHg). Persistent signal changes were observed after hypercapnia. These were compared to both baseline and hypercapnia by taking the average of the signal from 7 to 9 min. Only the post-hypercapnia ΔoxCCO measured at rSD = 3 cm was significantly lower than its corresponding hypercapnic value ( Table 2).  Figure 7 presents average time courses of changes in BFi, StO 2 , and oxCCO in response to 4 min of hypercapnia (P ET CO 2 increase = 10 ± 2 mmHg). Hypercapnic responses were calculated as the relative difference between the average signal from 4 to 6 min (i.e., 2nd half of the hypercapnic period) and the average of the first minute of baseline. Average ∆oxCCO and ∆StO 2 recorded at both source-detectors distances are provided in Table 2. ∆oxCCO and ∆StO 2 measured at r SD = 3 cm were significantly larger than the responses measured at r SD = 1 cm. The ∆oxCCO and ∆StO 2 responses at 1 cm were significantly delayed (33 ± 24 s and 10 ± 6 s, respectively) compared to the corresponding responses at 3 cm. Hypercapnia resulted in a significant increase in BFi (31 ± 48%) and MAP (4 ± 1 mmHg). Persistent signal changes were observed after hypercapnia. These were compared to both baseline and hypercapnia by taking the average of the signal from 7 to 9 min. Only the post-hypercapnia ∆oxCCO measured at r SD = 3 cm was significantly lower than its corresponding hypercapnic value (Table 2).     (Table 2); however, their responses were not significantly different from the original responses measured at rSD = 3 cm. Posthypercapnia, ΔStO2,Reg returned to baseline and was significantly smaller than the corresponding post-hypercapnia ΔStO2,3cm.    (Table 2); however, their responses were not significantly different from the original responses measured at r SD = 3 cm. Post-hypercapnia, ∆StO 2,Reg returned to baseline and was significantly smaller than the corresponding posthypercapnia ∆StO 2,3cm .

Discussion
This study focused on evaluating contributions from the scalp and brain on metabolic and hemodynamic markers measured with hsNIRS. The primary motivation was to improve the confidence in non-invasive ΔoxCCO monitoring for cardiac and vascular surgery applications. In this study, the impact of the scalp was assessed by comparing signals measured at rSD = 1 cm, which predominately represents changes in the extracerebral layer, and rSD = 3 cm, which contains a greater brain contribution. The study involved two paradigms: unilateral CC and hypercapnia. The motivation for using CC was that it is a safe method of causing rapid and large decreases in cerebral blood flow that mimics arterial occlusion performed during surgery. Hypercapnia was included given its well-known vasodilatory effects in the brain.
The average reductions in BFi for 15 and 30-s periods of CC were 55 ± 8% and 57 ± 14%, respectively, which are consistent with a 60% decrease in mean blood flow velocity measured in the middle cerebral artery by transcranial Doppler [30]. Repeat 15-s CC trials demonstrated that ΔoxCCO measured at both source-detector distances was highly reproducible with a CoV of 6% at rSD = 1 cm and 1% at 3 cm. Thirty seconds of CC decreased oxCCO by 0.4 ± 0.3 μM at rSD = 3 cm (Figure 3a). The magnitude of this decrease is greater than reported for other experimental paradigms, including mild hypoxia, hypocapnia [28], and breath holding [49]. More importantly, the average oxCCO reduction at rSD = 3 cm was almost seven times greater than the corresponding oxCCO reduction measured at rSD = 1 cm (0.06 ± 0.1 μM). The latter was not significantly different from the baseline. The significant difference in the oxCCO responses at the two distances (p = 0.012, ΔoxCCO at rSD = 3 vs. 1 cm) reflects the greater brain contribution to the signal measured at rSD = 3 cm. Considering the higher metabolic rate of the brain compared to scalp and the higher cerebral oxCCO concentration, a sudden and sizable decrease in oxygen delivery would likely have a greater effect on the brain. The greater sensitivity of the oxCCO signal to the brain is exemplified in Figure 5, which shows that changes in oxCCO measured at rSD = 1 cm never reached significance across all BFi decreases; whereas, changes in oxCCO at rSD = 3 cm were significant for all decreases in BFi greater than 20%.
The hypercapnia results also demonstrated the sensitivity of the oxCCO signal to the brain. The average increase in PETCO2 was 10 ± 2 mmHg, which caused a 31 ± 48% increase in BFi and a significant increase in oxCCO of 0.22 ± 0.19 μM measured at rSD = 3 cm (Table  2). Similar to the CC results, the oxCCO change measured at rSD = 1 cm (0.1 ± 0.1 μM) did

Discussion
This study focused on evaluating contributions from the scalp and brain on metabolic and hemodynamic markers measured with hsNIRS. The primary motivation was to improve the confidence in non-invasive ∆oxCCO monitoring for cardiac and vascular surgery applications. In this study, the impact of the scalp was assessed by comparing signals measured at r SD = 1 cm, which predominately represents changes in the extracerebral layer, and r SD = 3 cm, which contains a greater brain contribution. The study involved two paradigms: unilateral CC and hypercapnia. The motivation for using CC was that it is a safe method of causing rapid and large decreases in cerebral blood flow that mimics arterial occlusion performed during surgery. Hypercapnia was included given its well-known vasodilatory effects in the brain.
The average reductions in BFi for 15 and 30-s periods of CC were 55 ± 8% and 57 ± 14%, respectively, which are consistent with a 60% decrease in mean blood flow velocity measured in the middle cerebral artery by transcranial Doppler [30]. Repeat 15-s CC trials demonstrated that ∆oxCCO measured at both source-detector distances was highly reproducible with a CoV of 6% at r SD = 1 cm and 1% at 3 cm. Thirty seconds of CC decreased oxCCO by 0.4 ± 0.3 µM at r SD = 3 cm (Figure 3a). The magnitude of this decrease is greater than reported for other experimental paradigms, including mild hypoxia, hypocapnia [28], and breath holding [49]. More importantly, the average oxCCO reduction at r SD = 3 cm was almost seven times greater than the corresponding oxCCO reduction measured at r SD = 1 cm (0.06 ± 0.1 µM). The latter was not significantly different from the baseline. The significant difference in the oxCCO responses at the two distances (p = 0.012, ∆oxCCO at r SD = 3 vs. 1 cm) reflects the greater brain contribution to the signal measured at r SD = 3 cm. Considering the higher metabolic rate of the brain compared to scalp and the higher cerebral oxCCO concentration, a sudden and sizable decrease in oxygen delivery would likely have a greater effect on the brain. The greater sensitivity of the oxCCO signal to the brain is exemplified in Figure 5, which shows that changes in oxCCO measured at r SD = 1 cm never reached significance across all BFi decreases; whereas, changes in oxCCO at r SD = 3 cm were significant for all decreases in BFi greater than 20%.
The hypercapnia results also demonstrated the sensitivity of the oxCCO signal to the brain. The average increase in P ET CO 2 was 10 ± 2 mmHg, which caused a 31 ± 48% increase in BFi and a significant increase in oxCCO of 0.22 ± 0.19 µM measured at r SD = 3 cm (Table 2). Similar to the CC results, the oxCCO change measured at r SD = 1 cm (0.1 ± 0.1 µM) did not reach significance. This finding is in agreement with Kolyva et al., who reported that the magnitude of the oxCCO response to hypercapnia increased with source-detector separation (2 ≤ r SD ≤ 3.5 cm) [32]. Note, there is some debate as to whether oxCCO should increase during hypercapnia if it is close to fully oxidized at normoxia [50,51]. The consistent increase in oxCCO observed in human participants indicates that this is likely not the case [32].
This study also demonstrated that StO 2 was more sensitive to extracerebral tissue than oxCCO. Similar to oxCCO, greater changes in StO 2 were measured at r SD = 3 cm compared to 1 cm for CC; however, the ratio of ∆StO 2 measured at the two distances was around three, in contrast to a ratio closer to seven for ∆oxCCO. Moreover, unlike oxCCO, there was a significant decrease in StO 2 measured at r SD = 1 cm (p = 0.001 vs. baseline) ( Table 1). The hemodynamic response of ∆StO 2 to CC was also significantly slower, as characterized by the time constant τ, which was larger for ∆StO 2 compared to the corresponding values for ∆oxCCO and ∆BFi. The average time courses for the three parameters ( Figure 3) demonstrated that ∆oxCCO followed ∆BFi more closely than ∆StO 2 . The StO 2 response likely reflects a slower response to CC in the metabolically inactive scalp tissue. In newborn piglets, which have thin skulls and negligible scalp muscle, Rajaram et al. observed that CBF and StO 2 both decreased rapidly in response to hypoxia-ischemia while oxCCO displayed a delayed response [40]. The use of hypoxia in the piglet study may also have contributed to the difference between these two studies since SaO 2 was not altered in the CC experiments.
A further illustration of the sensitivity of StO 2 to the extracerebral tissue was the persistent elevation observed after hypercapnia (Figure 7). In healthy participants, cerebrovascular reactivity will be reflected by rapid changes in StO 2 at the onset and end of hypercapnia, as demonstrated in functional magnetic resonance imaging studies [48]. The influence of the scalp, which has considerably more sluggish vascular reactivity, was previously demonstrated using time-resolved NIRS. Only hemoglobin signals with enhanced depth sensitivity exhibited a rapid return to baseline when P ET CO 2 returned to normocapnia [33,35]. In the current study, ∆StO 2 at r SD = 3 cm remained significantly greater than baseline (p = 0.003) one to three minutes after hypercapnia. In contrast, ∆oxCCO at r SD = 3 cm in the same period was not significantly different from baseline. However, some evidence of scalp contamination in oxCCO measurements was also observed. Post hypercapnia, ∆oxCCO at r SD = 1 cm was significantly greater than at baseline, reflecting some sensitivity to the scalp (Table 2).
Regression analysis was explored as a means of reducing scalp contamination in the hsNIRS data. For both ∆oxCCO and ∆StO 2 during CC, regression reduced the magnitude and inter-subject variability ( Figure 4); however, these changes were not significant. More apparent effects can be observed in the regression results obtained for hypercapnia ( Figure 8). For the hypercapnia data, regression did significantly reduce post hypercapnia ∆StO 2 (p = 0.02 vs. ∆StO 2 at 3 cm) ( Table 2). A similar effect can be observed for ∆oxCCO; however, the signal change did not reach significance. This effect suggests that scalp contamination also affected the ∆oxCCO signal at r SD = 3 cm but to a lesser extent than ∆StO 2 .
This study presented a few limitations. The ratio of the CC responses at the two sourcedetector separations (Table 1) was larger for ∆oxCCO than StO 2 , but this difference was not significant. Power analysis indicated that 42 participants would have been required to show significance. Next, only a single source-detector separation was used for DCS acquisition. While the DCS will include contributions from scalp blood flow, the magnitude of this contamination will be less compared to NIRS due to the higher blood flow in the brain. Since this study was primarily focused on hsNIRS measurements of oxCCO, single-distance DCS measurements were deemed reasonable. Finally, regression analysis is sensitive to signal noise and cannot be performed in real-time in clinical settings. Future work will focus on assessing real-time methods for reducing scalp contributions from multi-distance hsNIRS data.

Conclusions
In summary, the study measured oxCCO and StO 2 changes with hsNIRS at multiple source-detector distances during two paradigms. The first paradigm, CC, caused substantial reductions in blood flow, analogous to hemodynamic events that can occur during cardiac and vascular surgeries. The novelty of this study was demonstrating a significant decrease in oxCCO at r SD = 3 cm during CC but not at r SD = 1 cm. In contrast, significant decreases in StO 2 were observed at both distances. These results indicate that oxCCO had less scalp contamination than concurrent StO 2 measurements. These results highlight the potential of using oxCCO to monitor brain health during surgery. However, increases in oxCCO were observed in the post hypercapnia data acquired at r SD = 1 cm, indicating some contamination from the scalp. Therefore, acquiring multi-distance hsNIRS data and applying methods to separate scalp and brain contributions, such as regression analysis, is likely prudent for interpreting changes in oxCCO in clinical settings.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the Review Board of Western University, which adheres to the guidelines of the Tri-Council Policy Statement for research involving humans (REB # 120391 and 17 January 2021).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Data can be made available by contacting the authors. Because of the participant consent obtained as part of the recruitment process, it is not possible to make these data publicly available.