2.1.

Participants

All procedures were approved by the Health Sciences Research Ethics Board (HSREB) of Western University (No. 105417) and conducted in accordance with the Declaration of Helsinki ethical standards. Participants provided written informed consent following verbal and written explanations of the experimental procedures. Thirteen healthy participants were recruited. Before instrumentation, participants completed a mandatory health history form to evaluate the inclusion/exclusion criteria. Participants were included if they passed the MRI screen form and reported being nonsmokers with no prior diagnosis of cardiovascular disease, neurological disorder, diabetes, or hypertension. All subjects reported that they identify with their sex assigned at birth.

One male participant was excluded from all analysis as a result of equipment difficulties. Therefore, a total of twelve participants ($n=12$) were subsequently included in the ASL and DCS hypercapnia analysis. A subset of participants ($n=9$; 4 female) completed two additional protocols (DHC-NIRS and hypercapnia with trNIRS) described below.

2.2.

Instrumentation

Participants breathed through a mask sealed to the face with skin tape (Tegaderm, 3M, Saint Paul, MN) to prevent air leakage. The facemask was connected to a computerized sequential gas delivery system (RespirAct™, Thornhill Medical, Toronto, Canada) that controlled end-tidal partial pressure of oxygen (${\mathrm{P}}_{\mathrm{ET}}{\mathrm{O}}_2$) and carbon dioxide (${\mathrm{P}}_{\mathrm{ET}}{\mathrm{CO}}_2$) independent of the pattern of participant ventilation.

A custom-designed optode holder was placed on the subject’s right lateral forehead and secured with a Velcro headband. Importantly, the headband was tightened to reduce extracerebral blood flow contamination with optical measures.^30–32

The study utilized a custom-built hybrid system that combined trNIRS and DCS instruments, which is described in detail elsewhere.^28,33–35 Briefly, the trNIRS module used two pulsed lasers (repetition rate = 80 MHz) operating at 760 and 832 nm (PicoQuant, Berlin, Germany). Light pulses from the laser heads were coupled into a multimode bifurcated fiber ($\mathrm{\Phi }=200\mu \mathrm{m}$, $\mathrm{NA}=0.22$, Loptek, Germany). Diffusely reflected light was collected at a source-detector separation ($r_{\mathrm{SD}}$) of 3 cm and delivered to a hybrid photomultiplier tube (PMA Hybrid 50, PicoQuant, Berlin, Germany) via a multimodal fiber ($\mathrm{\Phi }=400\mu \mathrm{m}$, $\mathrm{NA}=0.39$, Loptek, Germany). A time-correlated single-photon counting unit (HydraHarp 400, PicoQuant, Berlin, Germany) was used to record photon arrival times and build distributions of time-of-flight (DTOFs³⁶). To enable simultaneous trNIRS/DCS acquisition, a short-pass interference filter (Spec 3551, 836.5 nm, diameter 25 mm, Alluxa, Santa Rosa, California, United States) was placed in front of the trNIRS detector. Finally, the system instrument response function (IRF) was measured at the end of each experiment using a custom light-tight box.

The DCS module employed a long coherence length laser operating at 852 nm (CrystaLaser, Reno, Nevada, United States). The laser light was delivered to the tissue through a multimode fiber ($\mathrm{\Phi }=400\mu \mathrm{m}$, $\mathrm{NA}=0.39$, FT400UMT, Thorlabs, Newton, New Jersey, United States). The diffusively reflected light was collected by three single-mode fibers ($\mathrm{\Phi }=4.4\mathrm{nm}$, $\mathrm{NA}=0.13$, 780HP, Operating Wavelength = 780 to 970 nm, Thorlabs) at $r_{\mathrm{SD}}=2.7\mathrm{cm}$, which were coupled to a four-channel single photon counting module (SPCM-AQR-15-FC, Excelitas Technologies, Montreal, QC, Canada). The output from each detector was fed into an edge-detecting counter on a PCIe-6612 counter/timer data acquisition board (National Instruments, Austin, Texas, United States). Photon counts were recorded and processed using custom software (LabVIEW, National Instrument, Austin, Texas, United States³⁷) that generated intensity autocorrelation curves for 50 delay times ($\tau$) ranging from $1\mu \mathrm{s}$ to 1 ms.³⁸ Time-resolved NIRS and DCS data were acquired at a sampling frequency of 3 Hz in all protocols.

2.3.

Experimental Design

After obtaining informed consent and completing instrumentation (described above), participants provided a venous blood sample (1 mL) from the antecubital vein to assess baseline hematocrit concentration (ABL80 Flex Co-ox, Radiometer, Copenhagen, Denmark). All protocols consisted of participants resting in a supine posture. The headband and optode holder were not moved between protocols. Briefly, participants completed three protocols, listed in order of occurrence: (1) ASL hypercapnia protocol, (2) dynamic hypoxia contrast (DHC) protocol, and (3) trNIRS hypercapnia protocol.

2.4.

MRI Acquisition

Hypercapnia experiments were performed using a 3T Biograph mMR scanner (Siemens Medical Systems, DE, Erlangen, Germany) with a 12-channel receive-only head coil. The head was immobilized with foam padding to minimize motion artifacts, and MRI fiducial marks were placed on the forehead to identify the location of the optodes on structural MRIs.

Sagittal T1-weighted images were acquired using a 3D magnetization-prepared rapid gradient echo sequence [MPRAGE; repetition time (TR) = 2 s, echo time (TE) = 2.98 ms, inversion time = 900 ms, flip angle = 9 deg, field of view $(\mathrm{FOV})=176\times 256\times 256\mathrm{mm}$, and isotropic $\text{voxel size}=1.0{\mathrm{mm}}^3$]. CBF images were obtained using a pseudo-continuous arterial spin labeling (pCASL) sequence with 3D segmented gradient and spin-echo readout that incorporated background suppression (voxel size $3.8\times 3.8\times 3.8{\mathrm{mm}}^3$; 26 slices; TR = 4.22 s; TE = 39.76 ms; single post labeling delay = 1.7 s; labeling duration =1.5 s; $\mathrm{FOV}=240\times 240\times 99\mathrm{mm}$).³⁹ A total of 48 pairs of tagged and control images were obtained at rest and during hypercapnia for each subject, along with a proton density-weighted image M0 (no background suppression, TR = 10 s).

2.5.

ASL and trNIRS Hypercapnia Protocols

The ASL hypercapnia protocol consisted of 800 s ($\sim 13\min$) of simultaneous pCASL and DCS data acquisition. Datasets were collected during 409 s of normocapnic baseline, 323 s of hypercapnia (step increase of $+10\mathrm{mmHg}$), and 68 s of recovery. Cerebrovascular reactivity to ${\mathrm{CO}}_2$ (${\mathrm{CVR}}_{\mathrm{CO}2}$) was calculated as $\mathrm{\Delta }{\mathrm{CBF}}_{\mathrm{ASL}}/\mathrm{\Delta }{\mathrm{P}}_{\mathrm{ET}}{\mathrm{CO}}_2$ (mL/100 g/min/mmHg). Upon completion of the MRI hypercapnia protocol, participants were transferred from the MRI to a hospital bed ($\sim 3\mathrm{m}$ walking distance) and given 10 min of quiet, supine rest. Without moving the optode holder, the DCS fibers were replaced with the trNIRS fibers. After completion of the DHC protocol (described below), participants were given 10 min of wash-out prior to starting the trNIRS hypercapnia protocol. The trNIRS hypercapnia protocol consisted of a 60-s normocapnic baseline, 5 min of hypercapnia (step increase of $+10\mathrm{mmHg}$), and 120 s of recovery. Importantly, ${\mathrm{P}}_{\mathrm{ET}}{\mathrm{CO}}_2$ was controlled by the RespirAct™ to ensure repeatability between protocols.

2.6.

Dynamic Hypoxia Contrast Protocol

The DHC protocol occurred under the same conditions as discussed above and immediately prior to the trNIRS hypercapnia protocol. This programmed hypoxia protocol was 180 s long (e.g., Fig. 1), consisting of a 60-s baseline ${\mathrm{P}}_{\mathrm{ET}}{\mathrm{O}}_2$ of 95 mmHg (normoxia), a step decrease in ${\mathrm{P}}_{\mathrm{ET}}{\mathrm{O}}_2$ to 40 mmHg (hypoxia) for 60 s, and a return to normoxia for the remaining 60 s.

Fig. 1

Average time courses of (a) arterial oxygen saturation (${\mathrm{SaO}}_2$) and (b) change in oxyhemoglobin concentration in the brain ($\mathrm{\Delta }{C}_{\mathrm{HbO}}$) ($n=9$). Shading around each line represents the standard deviation. (c) $\mathrm{\Delta }{C}_T$ time series of experimental data from a representative participant, along with the best fit of the perfusion model [Eq. (3)].

3.1.

MRI Image Analysis

Analysis was conducted with in-house Matrix Laboratory (MATLAB) scripts. Using the Statistical Parametric Mapping software package (SPM12),⁴⁰ MPRAGE images were segmented into tissue probability maps for grey and white matter, cerebrospinal fluid, and non-brain matter. These images were co-registered to the pCASL M0 image.

All pCASL images were realigned to M0 for motion correction. Perfusion-weighted images were generated by averaging the pairwise subtractions of sequential tag and control images, then calibrated to M0 for CBF quantification using the equation from Ref. 39.

Finally, each participant’s MPRAGE was used to measure the extracerebral layer thickness (ImageJ 1.54d). Five measurements of each layer thickness were made directly posterior to the DCS fiducial marker and averaged to yield mean skull and scalp layer thicknesses for each participant.

3.2.

DCS Analysis

Using the Siegert relation, each normalized intensity autocorrelation function (${\mathrm{g}}_2$) was converted to the corresponding electric field autocorrelation function and fit with the solution to the diffusion approximation for a semi-infinite homogenous medium.¹⁴ The fitting incorporated subject-specific optical coefficients obtained by trNIRS and the coherence factor ($\beta$) determined from the average initial value of the baseline $g_2$ curve. The fitting procedure was performed for the initial part of the intensity autocorrelation curve, i.e., $g_2(\tau )>1.25$, to increase the sensitivity to the brain. The fitting procedure yielded a best-fit blood flow index (BFi) estimate based on modeling tissue perfusion as pseudo-Brownian motion.⁶ The resulting BFi time courses were smoothed with the zero-phase digital filter (filtfilt, MATLAB, 2016b, MathWorks, Natick, Massachusetts, United States).

3.3.

Time-Resolved NIRS Analysis

Baseline subject-specific optical properties at each wavelength were determined from a DTOF averaged across the first 30 s of data. The mean DTOF was fit with the solution to the diffusion equation for a semi-infinite homogeneous medium convolved with the measured IRF (fminsearch, MATLAB, Mathworks Inc., Natick, Massachusetts, United States).³³ The fitting parameters were the absorption coefficient (${\mu }_a$), the reduced scattering coefficient (${\mu }_s^{\prime }$), and an amplitude factor that accounts for laser power, detection gain, and coupling efficiency. The fitting range was set to 10% of the peak value of a DTOF on the leading edge and 5% on the falling edge.²⁹

The first three statistical moments (number of photons $N$, mean time of flight $⟨t⟩$, and variance $V$) were used to obtain depth sensitivity from the trNIRS data. The three moments were calculated for each DTOF in a time series recorded at either 760 or 832 nm by setting the lower and upper integration limits based on arrival times corresponding to 1% of the peak of the DTOF. The change in each moment relative to its initial value (mean value across 30 s at baseline) was calculated to generate three time series (i.e., $\mathrm{\Delta }N$, $\mathrm{\Delta }⟨t⟩$, and $\mathrm{\Delta }V$), with $\mathrm{\Delta }N$ being more representative of extracerebral tissue and $\mathrm{\Delta }V$ being more representative of cerebral tissue. The time courses determined for each moment were converted into absorption changes $\mathrm{\Delta }{\mu }_a(\lambda )$ using sensitivity analysis, as described previously.²⁹ The $\mathrm{\Delta }{\mu }_a(\lambda )$ time courses for 760 and 832 nm were converted to changes in concentration of oxyhemoglobin ($\mathrm{\Delta }{C}_{\mathrm{HbO}}$) and deoxy-hemoglobin ($\mathrm{\Delta }{C}_{\mathrm{Hb}}$) using wavelength-specific molar extinction coefficients.⁴¹

3.4.

Dynamic Hypoxia Contrast Analysis

As ${\mathrm{P}}_{\mathrm{ET}}{\mathrm{O}}_2$ is a valid index of arterial partial pressures of ${\mathrm{O}}_2$,²⁵ the ${\mathrm{P}}_{\mathrm{ET}}{\mathrm{O}}_2$ time course recorded by the RespirAct was converted to ${\mathrm{SaO}}_2$ using the Hill equation to describe the in vivo oxyhemoglobin dissociation curve.⁴² Arterial concentration of oxy-hemoglobin, $C_a(t)$, was calculated from the change in ${\mathrm{SaO}}_2$, denoted $\mathrm{\Delta }{S}_a{O}_2$, and subject-specific total hemoglobin concentration, tHb, using the following equation:

Baseline CBF (${\mathrm{CBF}}_{\mathrm{DHC}}$) was estimated by modeling the relationship between $C_a(t)$ and the corresponding time-varying change in tissue hemoglobin concentration, $C_T(t)$, as a linear time-invariant system assuming a constant blood flow:²¹

3.5.

Statistical Analysis

Individual normocapnic and hypercapnic values were calculated as the average value from a $\ge 2\text{-}\min$ steady-state response profile. All data are presented as mean ± standard deviation unless otherwise noted. Statistical significance was defined as $p<0.05$. All datasets were found to conform to a normal distribution based on visual inspection of $Q\text{-}Q$ plots and the results of the Shapiro-Wilk test (i.e., all $p\ge 0.22$). Sex differences between normocapnic ${\mathrm{CBF}}_{\mathrm{ASL}}$ and ${\mathrm{CVR}}_{\mathrm{CO}2}$ were assessed using two-tailed $t$-tests. Paired $t$-tests were used to assess ${\mathrm{CBF}}_{\mathrm{ASL}}$ differences between normo- and hypercapnia. A two-way analysis of variance was used to compare moments (3 levels: $\mathrm{\Delta }N$, $\mathrm{\Delta }⟨t⟩$, and $\mathrm{\Delta }V$) across time (16 levels: 30-s bins) with post hoc testing completed using Dunnett’s test for multiple comparisons (GraphPad Prism version 10.1.2). We assessed the average $\mathrm{\Delta }{\mu }_a$ from 832 nm ($\mathrm{\Delta }⟨t⟩$) as the difference between steady-state hypercapnia (minimum 2 min) and baseline values.

To examine the correlation between two techniques of deriving CBF, linear regression analysis (i.e., $R^2$ and $r$) was performed for the following data: [1] ${\mathrm{CBF}}_{\mathrm{DHC}}$ and ${\mathrm{CBF}}_{\mathrm{ASL}}$ ($\mathrm{mL}/100\mathrm{g}/\min$), [2] $\mathrm{\Delta }{\mathrm{CBF}}_{\mathrm{DCS}}$ and $\mathrm{\Delta }{\mathrm{CBF}}_{\mathrm{ASL}}$ (%), and [3] ${\mathrm{CBF}}_{\mathrm{DCS}}$ and ${\mathrm{CBF}}_{\mathrm{ASL}}$ ($\mathrm{mL}/100\mathrm{g}/\min$). The slope for each regression was statistically compared to a slope of 1. In addition, Bland–Altman analysis was conducted to assess the similarity between the two sets of CBF measurements.

4.1.

Absolute CBF from DHC-NIRS

The average reduction in ${\mathrm{SaO}}_2$ during transient hypoxia and the corresponding average time course of change in oxyhemoglobin concentration are shown in Fig. 1. The former was obtained from the change in ${\mathrm{P}}_{\mathrm{ET}}{\mathrm{O}}_2$ measured by the RespirAct and the latter from $\mathrm{\Delta }⟨t⟩$. Included in the figure is a representative case showing the perfusion model [Eq. (3)] fit to oxyhemoglobin time course from one subject.

Average ${\mathrm{CBF}}_{\mathrm{DHC}}$ was $44\pm 9\mathrm{mL}/100\mathrm{g}/\min$ ($n=9$) and $t_c$ was $8.49\pm 3.10\mathrm{s}$. There was a strong positive correlation between absolute baseline ${\mathrm{CBF}}_{\mathrm{ASL}}$ and ${\mathrm{CBF}}_{\mathrm{DHC}}$ ($r=0.77$, $p=0.02$; Fig. 2), with a slope of 0.79 that was not significantly different than 1.00 ($p=0.34$) and an intercept not significantly different than 0.00 ($p=0.06$). However, ${\mathrm{CBF}}_{\mathrm{DHC}}$ slightly underestimated ${\mathrm{CBF}}_{\mathrm{ASL}}$, as demonstrated by a mean bias of $-3\mathrm{mL}/100\mathrm{g}/\min$ with moderately wide limits of agreement [95% CI: $-15$, 8] and no evident trend or inconsistent variability.

Fig. 2

Regression and Bland–Altman plots comparing baseline CBF measures from arterial spin labeling CBF (${\mathrm{CBF}}_{\mathrm{ASL}}$) and dynamic hypoxia contrast time-resolved NIRS (${\mathrm{CBF}}_{\mathrm{DHC}}$) ($n=9$). The mean difference between the two methods is indicated by the solid black line, which was bound by a 95% confidence interval indicated by the dashed black lines.

4.2.

CBF_ASL during ASL Hypercapnia Protocol

Figure 3 presents absolute CBF images acquired with pCASL during normocapnia and hypercapnia ($n=12$). Overall, ${\mathrm{CBF}}_{\mathrm{ASL}}$ increased by $38\pm 9\%$ ($68\pm 11\mathrm{mL}/100\mathrm{g}/\min$; $p<0.01$ versus baseline) during hypercapnia ($9\pm 1\mathrm{mmHg}$ ${\mathrm{P}}_{\mathrm{ET}}{\mathrm{CO}}_2$). The average ${\mathrm{CVR}}_{\mathrm{CO}2}$ was $4.1\pm 1.0\%/\mathrm{mmHg}$, with no difference between males and females ($4\pm 1$ versus $4\pm 0\%/\mathrm{mmHg}$; $p=0.42$).

Fig. 3

Group average ($n=12$) ASL images acquired at baseline (normocapnia) and hypercapnia. The voxel-wise map of cerebrovascular reactivity was calculated as the perfusion change from normocapnia to hypercapnia, divided by 9 mmHg. The bottom row shows corresponding T1-weighted MR images for anatomical reference.

4.3.

ΔC_Hb and ΔC_HbO during trNIRS Hypercapnia Protocol

The average time courses of $\mathrm{\Delta }{C}_{\text{HbO}}$ and $\mathrm{\Delta }{C}_{\mathrm{Hb}}$ during hypercapnia are illustrated in Fig. 4. These time courses were derived from moment analysis of the trNIRS data. Significant interaction effects (moment-by-time) revealed differences between the moments during hypercapnia for both $\mathrm{\Delta }{C}_{\mathrm{HbO}}$ and $\mathrm{\Delta }{C}_{\mathrm{Hb}}$ (both $p<0.01$). For example, the $\mathrm{\Delta }{C}_{\mathrm{HbO}}$ obtained with ΔV was significantly greater than that calculated using $\mathrm{\Delta }N$ between 240 and 300 s (both $p\le 0.049$). However, there were no differences when comparing $\mathrm{\Delta }⟨t⟩$ to $\mathrm{\Delta }N$ or $\mathrm{\Delta }V$ (all $p\ge 0.07$). The reduction in $\mathrm{\Delta }{C}_{\mathrm{Hb}}$ calculated by $\mathrm{\Delta }⟨t⟩$ was greater than for $\mathrm{\Delta }N$ between 210 and 300 s (all $p\le 0.04$). However, there was no difference between outcomes from $\mathrm{\Delta }⟨t⟩$ and $\mathrm{\Delta }V$ (all $p\ge 0.28$). Importantly, the recovery to baseline was similar between all moments (all $p\ge 0.28$). $\mathrm{\Delta }{\mu }_a$ increased by $4.57\pm 2.26\%$ during hypercapnia.

Fig. 4

Average time courses of concentration changes in oxyhemoglobin $\mathrm{\Delta }{C}_{\mathrm{HbO}}$ and deoxy-hemoglobin $\mathrm{\Delta }{C}_{\mathrm{Hb}}$ as derived from the first three statistical moments (number of photons $N$, mean time of flight $⟨t⟩$, and variance $V$) throughout hypercapnia protocol ($+9\mathrm{mmHg}{\mathrm{P}}_{\mathrm{ET}}{\mathrm{CO}}_2$; grey rectangle). Shading around each line represents the standard deviation ($n=9$).

4.4.

CBF_ASL Versus BFi during ASL Hypercapnia Protocol

Figure 5 presents the average relative change CBF during hypercapnia as measured by ASL (grey matter) and DCS. BFi increased by $27\pm 12\%$ ($p<0.01$ versus baseline), which was not significantly different from ${\mathrm{CBF}}_{\mathrm{ASL}}$ ($p=0.05$). There was a strong positive and significant correlation between the change in ${\mathrm{CBF}}_{\mathrm{ASL}}$ and BFi ($r=0.68$, $p=0.01$; Fig. 6), with a slope of 0.88 that was not significantly different from 1.00 ($p=0.68$) and an intercept that was significantly lower than 0.00 ($p=0.04$). BFi slightly underestimated the increase in CBF measured by ASL, as demonstrated by a mean bias of $-5.6\%$ with moderately wide limits of agreement [95% CI: $-23$, 12] and no evident trend or inconsistent variability.

Fig. 5

Average time courses of cerebral blood flow calculated using arterial spin labeling (${\mathrm{CBF}}_{\mathrm{ASL}}$) and DCS (BFi). Shading around each line represents the standard deviation ($n=12$). The grey rectangle indicates the hypercapnia period. During steady-state hypercapnia, ${\mathrm{CBF}}_{\mathrm{ASL}}$ increased by $38\pm 9\%$ and BFi increased by $27\pm 12\%$.

Fig. 6

Regression and Bland–Altman plots comparing the hypercapnic CBF change measured by arterial spin labeling (${\mathrm{CBF}}_{\mathrm{ASL}}$) and DCS (BFi) ($n=12$). The mean difference between the two methods is indicated by the solid black line, which was bound by a 95% confidence interval indicated by the dashed black lines.

4.5.

CBF_ASL Versus CBF_DCS

Baseline CBF from DHC-NIRS was used to convert the relative BFi time series from DCS into CBF units (i.e., ${\mathrm{CBF}}_{\mathrm{DCS}}$). An exemplary tracing of absolute CBF at rest and during hypercapnia is illustrated in Fig. 7. Moreover, applying this conversion to baseline and hypercapnia BFi values resulted in CBF values that were moderately strong and positively associated with ${\mathrm{CBF}}_{\mathrm{ASL}}$ (both $p=0.02$; Fig. 8).

Fig. 7

Exemplar CBF time series from one participant from DCS and ASL. For the former, baseline CBF measured by DHC-NIRS was used to scale the DCS data. The hypercapnia period ($+9\mathrm{mmHg}$) is indicated by the grey rectangle.

Fig. 8

Regression plot comparing CBF values from ASL (${\mathrm{CBF}}_{\mathrm{ASL}}$) and calibrated BFi (${\mathrm{CBF}}_{\mathrm{DCS}}$) during normocapnic (baseline) and hypercapnia ($n=9$).