Sage Journals: Discover world-class research

Abstract

Functional MRI (fMRI) can identify active foci in response to stimuli through BOLD signal fluctuations, which represent a complex interplay between blood flow and cerebral metabolic rate of oxygen (CMRO₂) changes. Calibrated fMRI can disentangle the underlying contributions, allowing quantification of the CMRO₂ response. Here, whole-brain venous oxygen saturation (Y_v) was computed alongside ASL-measured CBF and BOLD-weighted data to derive the calibration constant, M, using the proposed Y_v-based calibration. Data were collected from 10 subjects at 3T with a three-part interleaved sequence comprising background-suppressed 3D-pCASL, 2D BOLD-weighted, and single-slice dual-echo GRE (to measure Y_v via susceptometry-based oximetry) acquisitions while subjects breathed normocapnic/normoxic, hyperoxic, and hypercapnic gases, and during a motor task. M was computed via Y_v-based calibration from both hypercapnia and hyperoxia stimulus data, and results were compared to conventional hypercapnia or hyperoxia calibration methods. Mean M in gray matter did not significantly differ between calibration methods, ranging from 8.5 ± 2.8% (conventional hyperoxia calibration) to 11.7 ± 4.5% (Y_v-based calibration in response to hyperoxia), with hypercapnia-based M values between (p = 0.56). Relative CMRO₂ changes from finger tapping were computed from each M map. CMRO₂ increased by ∼20% in the motor cortex, and good agreement was observed between the conventional and proposed calibration methods.

Keywords

Arterial spin labeling BOLD calibrated fMRI cerebral metabolic rate of oxygen magnetic resonance imaging

Introduction

Measurement of the brain's metabolic response to a stimulus provides insight into the regional recruitment and activation of neurons. As the brain relies almost entirely on oxidative metabolism to meet its energy needs, the cerebral metabolic rate of oxygen (CMRO₂) can be used to characterize the overall metabolic requirements of the brain. Traditional blood-oxygen level dependent (BOLD) MRI is sensitive to changes in the effective transverse relaxation rate (R₂*) due to local changes in deoxyhemoglobin (dHb) content and concentration.¹ In response to neural activation, the BOLD signal ultimately changes due to a combination of increased local metabolism (e.g. increased [dHb]) and a surplus of local blood flow (which tends to decrease [dHb]). Thus, due to the competing processes, traditional BOLD yields patterns of activation but cannot on its own quantify the magnitude of the CMRO₂ response.²

Calibrated fMRI, on the other hand, can quantify relative CMRO₂ changes in response to a stimulus,^3,4 and, more recently, voxel-wise CMRO₂ at rest.^5–7 By measuring BOLD and cerebral blood flow (CBF) changes (via arterial spin labeling, ASL) in response to a global stimulus – the so-called calibration scan, the relationship between BOLD, CBF, and CMRO₂ is revealed, yielding the calibration constant, M, which represents the maximum possible BOLD signal change that would be achieved with complete washout of dHb. Subsequently, task-evoked changes in BOLD and CBF can be converted to relative CMRO₂ changes.^3,8,9 Thus, following the calibration scan, calibrated fMRI can quantify relative CMRO₂ responses with both high spatial and temporal resolution.

Most calibration methods use vasoactive stimuli such as hypercapnia,^3,4 hyperoxia,^10,11 carbogen,¹² or a combination thereof,^5–7,13 to determine M. Separate assumptions and calibration equations are used, depending on the choice of stimulus. Hypercapnia (HC, e.g. ∼5% CO₂ in room air) induces large blood flow changes in the cerebral cortex due to vasodilation and increased perfusion pressure secondary to changes in blood pressure.¹⁴ Hyperoxia (HO, increased fraction of inspired oxygen, FiO₂ > 21%) increases the oxygen carrying capacity of arterial blood via plasma-dissolved oxygen. More modest CBF changes are observed in response to HO, with reports suggesting that CBF remains relatively constant¹⁵ or slightly decreases.¹⁶ HO also induces a positive BOLD response^17,18 attributed to an overall decrease in local [dHb], as abundant plasma-dissolved oxygen reduces the amount of hemoglobin-bound O₂ offloaded to tissue.

Due to the relative paramagnetism of deoxyhemoglobin, MRI lends itself to the measurement of venous oxygen saturation (Y_v), where [dHb]∝(1-Y_v), through T₂-¹⁹ or susceptometry-based oximetry (SBO).^20,21 If it is assumed that changes in venous oxygen saturation are spatially uniform in response to the gas breathing stimulus, whole-brain measures of Y_v (e.g. from the superior sagittal sinus, SSS) may be used to better inform the calibration procedure. To that end, we herein propose a calibration method, termed Y_v-based calibration, applied identically to HO or HC stimuli, in which BOLD, ASL, and MR-measured whole-brain Y_v are incorporated to determine M. We introduce an interleaved ASL, BOLD, and dual-echo GRE sequence, OxBOLD, to concurrently measure CBF, BOLD, and Y_v. We compare M maps derived from conventional and proposed (Y_v-based) calibration models, and finally demonstrate the ability to measure the relative CMRO₂ response to finger tapping.

Theory

The goal of calibrated fMRI is to disentangle the CMRO₂ response from BOLD and CBF changes. As described by Davis, et al.,³ calibrated fMRI relates the BOLD signal changes to the relative response of the cerebral blood volume (CBV) and venous [dHb] as

\frac{Δ BOLD}{BOL D_{0}} = TE \cdot A \cdot CB V_{0} \cdot [dHb] v_{0} β \times (1 - (\frac{CBV}{CB V_{0}}) (\frac{[dHb] v}{[dHb] v_{0}}) β)

(1)

where

M = TE \cdot A \cdot CB V_{0} \cdot [dHb] v_{0} β

is the sequence (e.g. echo time, TE) and subject-specific calibration constant, representing the maximum BOLD signal change that would be achieved with full washout of dHb. The subscripts (0) represent the baseline condition relative to stimulus condition (no subscript), A a constant related to the main magnetic field (B₀) and the volume susceptibility difference between blood and tissue (Δχ), and β a B₀-dependent constant reflecting relative intra- and extra-vascular BOLD signal contributions. ΔBOLD is the stimulus-evoked BOLD signal change (BOLD-BOLD₀). By invoking Fick's principle, i.e.

CMR O_{2} \propto CBF (1 - Y_{v}

), and assuming that CBV and CBF are related through Grubb's constant,²² α, equation (1) can be recast to highlight the relationship between BOLD, CBF, and CMRO₂ through M as

\begin{matrix} \frac{Δ BOLD}{BOL D_{0}} = M (1 - (\frac{CMR O_{2}}{CMR O_{2} | 0}) β (\frac{CBF}{CB F_{0}}) α - β) \end{matrix}

(2)

There are a variety of strategies to solve for M, generally involving a calibration experiment in which BOLD, CBF, and sometimes [dHb]_v are simultaneously measured during a stimulus assumed to be isometabolic (typically HC or HO gas breathing). In the original model proposed by Davis et al.,³ using HC as a vasoactive stimulus with an assumed isometabolic response allows simplification of equation (2) to

\begin{matrix} \frac{Δ BOLD}{BOL D_{0}} = M (1 - (\frac{CBF}{CB F_{0}}) α - β) \end{matrix}

(3)

Alternatively, HO can be used as a stimulus for BOLD calibration.^10,11 Initially proposed by Chiarelli et al.,¹⁰ M can be determined by simultaneously measuring BOLD and CBF changes in response to HO, while estimating [dHb]_v from changes in the expired oxygen (e.g. partial pressure of end-tidal O₂, P_ETO₂). Given that [dHb]_v changes both because of increased oxygen content in arterial blood (proportional to the partial pressure of oxygen in inspired air) and changes in blood flow, a correction term is added to equation (1) as

\begin{matrix} \frac{Δ BOLD}{BOL D_{0}} \\ = M (1 - (\frac{CBF}{CB F_{0}}) α (\frac{[dHb] v}{[dHb] v_{0}} + \frac{CB F_{0}}{CBF} - 1) β) \end{matrix}

(4)

However, the conversion from changes in P_ETO₂ to [dHb] is rather indirect, relying on the assumption that oxygen extraction fraction (OEF) is constant between the normoxic and hyperoxic conditions, and that [dHb] changes are not only spatially uniform in the brain, but also throughout the body.

Alternatively, MRI can be used to quantify intravascular Y_v in a large draining vein using SBO. By modeling a vein, in this case the SSS, as a long, straight cylinder oriented relatively parallel to B₀, an analytical solution can relate the magnetic susceptibility-induced incremental field shift due to the presence of paramagnetic deoxyhemoglobin inside the vein (measured as the vein-tissue difference in the rate of inter-echo phase accrual from gradient echo data) to Y_v as^20,21

\begin{matrix} Y_{v} = [1 + \frac{Δ χ_{oxy}}{Δ χ_{do}} - \frac{2 Δ ϕ / Δ φ Δ TE Δ TE}{γ Δ χ_{do} Hct B_{0} (co s^{2} θ - 1 / 133)}] \times 100 \end{matrix}

(5)

where Δχ _do (=0.273 ppm, cgs) and Δχ_oxy (=−0.008 ppm, cgs) are the susceptibility differences between fully oxygenated and deoxygenated erythrocytes, and between fully oxygenated erythrocytes and water,^23,24 respectively. The difference of phase accrual between the vein and background reference tissue (Δϕ) is measured for a given inter-echo spacing (ΔTE), and γ is the gyromagnetic ratio, Hct is the hematocrit, and θ is the angle of the vein relative to B₀. By measuring Y_v in the SSS, the main draining vein of the brain, whole-brain oxygen extraction can be determined. If it is assumed that the stimulus condition (e.g. HO or HC) induces spatially uniform changes in Y_v, equation (1) can be recast as

\frac{Δ BOLD}{BOL D_{0}} = M (1 - (\frac{CBF}{CB F_{0}}) α (\frac{1 - Y_{v}}{1 - Y_{v} | 0}) β)

(6)

referred to as the Y_v-based calibration model.

With the Y_v-based calibration model, all parameters are obtained from the MRI experiment, and the brain's global oxygen extraction is directly measured, therefore it does not need to be assumed as constant between the baseline and stimulus conditions. Isometabolism does not need to be explicitly assumed; however, if no CMRO₂ change occurs in response to the stimulus, the Y_v-based model simplifies to the original model proposed by Davis et al.³ as shown in equation (2). Furthermore, the Y_v-based calibration model can theoretically be applied in the context of any global neurometabolic stimulus that produces spatially uniform changes in Y_v.

Materials and methods

Thirteen young healthy subjects (mean ± standard deviation age = 30.1 ± 4.2 years, 7 male) were recruited to participate in the study after providing written informed consent. All aspects of this study were approved by the Institutional Review Board of the University of Pennsylvania as guided by the ethical principles set forth in the Belmont Report. Each subject completed screening and medical history questionnaires, followed by MRI scanning procedures and a venous blood draw. OxBOLD (described below) data were collected for 5 min during each condition including (1) normocapnic/normoxic baseline, (2) HO, (3) finger tapping (while breathing normocapnic/normoxic air), and (4) HC. Partial pressure of end-tidal oxygen (P_ETO₂) and CO₂ (P_ETCO₂) were monitored and stimulus gases delivered via the RespirAct Gas Control System (Thornhill Research, Toronto, Canada). Following the MRI experiment, a venous blood draw was performed to determine the subject's hematocrit.

Image acquisition

Data were collected at 3T (Siemens Prisma, VE11C, Erlangen, Germany) using a 32-channel head coil. Subjects were positioned supine, and a mask was placed over their nose and mouth and secured with medical skin tape to achieve an air-tight seal. Localizers were followed by a time-of-flight angiographic sequence, used to plan the locations of the pseudo-continuous arterial spin labeling (pCASL) labeling plane and dual-echo GRE acquisition. Subsequently, a sagittal T₁-weighted MPRAGE was acquired for image registration and gray matter (GM) versus white matter (WM) segmentation. A series of OxBOLD scans, each lasting 5 min, were then collected at rest, while the subject was breathing stimulus gases (i.e. HO and HC), and while the subject performed a block-design finger tapping task.

OxBOLD is a three-part interleaved sequence comprising background-suppressed 3D-RARE stack-of-spirals pCASL,²⁵ BOLD-weighted 2D spiral, and dual-echo GRE acquisitions. To maximize the temporal SNR of both BOLD and ASL data, a dual-acquisition approach²⁶ was used. Briefly, background-suppression pulses were used to reduce the signal from static GM, WM, cerebrospinal fluid (CSF), and remnant local blood to ∼1% of the initial magnetization. Following the 3D-RARE stack-of-spirals readout, a 1.3 s delay was allotted for recovery of longitudinal magnetization prior to the BOLD-weighted acquisition. Following the full BOLD acquisition, a global saturation pulse was applied and single-slice, dual-echo GRE data were acquired to determine Y_v in the SSS via SBO.^20,21 The OxBOLD pulse sequence and longitudinal magnetization of various tissues in the brain over the course of one OxBOLD TR are shown in Figure 1.

Figure 1.

OxBOLD pulse sequence showing interleaved approach for ASL, BOLD, and susceptometry-based oximetry (SBO) data. Relative slice/slab locations are shown on the right, where the dark red block corresponds to the 3D ASL FOV with labeling plane is shown in green, the light red boxes show the BOLD-weighted slices, and the blue shows the single-slice SBO acquisition location. The relative longitudinal magnetization (Mz/M₀) over the course of one TR is shown on bottom for gray matter (GM, T₁ = 1300 ms), white matter (WM, T₁ = 800 ms), local blood (e.g. not accounting for blood flowing in from the labeling plane, T₁ = 1660 ms), and cerebrospinal fluid (CSF, T₁ = 4000 ms). The background-suppression pulses during the ASL labeling and post labeling delay periods reduce the static tissue signal to ∼1% of its initial value. A delay is allotted between the ASL and BOLD acquisitions and again between BOLD and SBO acquisitions to allow for signal recovery.

ASL and BOLD data were acquired with full-brain spatial coverage, albeit with 3D and 2D readouts, respectively. The pCASL labeling plane was positioned based on the time-of-flight acquisition at a location where the internal carotid and vertebral arteries were relatively straight and parallel to B₀. The dual-echo GRE data were acquired in the middle of the straightest portion of the SSS. Though the dual-echo GRE acquisition is single-slice, the measured data represent the brain's global average Y_v.

Acquisition parameters for each component of the OxBOLD sequence are as follows – pCASL: FOV = 240 × 240 × 96 mm³, 3D, single-shot, variable-density, spiral-out trajectory readout with duration = 265 ms, TE = 15.9 ms, reconstruction matrix = 64 × 64×16, unbalanced labeling with label duration = 1.8 s, post labeling delay (PLD)=1.5 s, and background suppression achieved with four successive pulses to reduce the background signal to 1% of its initial value as described in Maleki et al.²⁷ BOLD-weighted acquisition with: FOV = 240 × 240 mm², slice thickness = 6 mm, 16 slices, 2D, single-shot, spiral-out trajectory readout with reconstruction matrix = 64 × 64, and TE = 29 ms. Single-slice multi-echo GRE data were acquired near the straightest portion of the SSS with: FOV = 208 × 208 mm², slice thickness = 5 mm, acquisition matrix = 208 × 52, keyhole²⁸ reconstruction using fully-phase encoded reference scan obtained at the start of the acquisition yielding image matrix = 208 × 208, TR/TE/ΔTE=15/3.78/3.52 ms, and flip angle = 15°. Overall sequence TR=7 s, corresponding to one ASL tag-control pair every 14 s.

Gas manipulations

Gases were delivered and monitored via the RespirAct Gas Control System, which is capable of independently targeting P_ETO₂ and P_ETCO₂.²⁹ Subjects were connected to the system throughout the imaging experiment, breathing normoxic/normocapnic air during the localizers, T₁-MPRAGE, and baseline and finger tapping OxBOLD acquisitions. HO was prescribed as ΔP_ETO₂ = 230 mmHg, corresponding to FiO₂ ≈ 50%, and HC as ΔP_ETCO₂ = 8 mmHg. The transition between stimulus states was a step for HO and a 30 s ramp for HC. In all cases, the subject was given time to adjust to the stimulus (∼1 min) prior to the start of image acquisition.

Motor stimulus

To make efficient use of the scan session, motor stimulus data were collected between the HO and HC stimulus experiments. A block-design, bilateral finger tapping task was performed while the subjects breathed normocapnic/normoxic air. For 70 s blocks (equivalent to 10 TRs), subjects were instructed to rest, then tap each finger to their thumb serially, then rest, then tap. Subjects verbally confirmed their participation at the end of the scan.

Data processing and analysis

Data preprocessing and analysis were carried out using SPM (version 12, Wellcome Trust Center for Neuroimaging, University College London, UK) and custom scripts in Matlab (Mathworks, MA).

The T₁ anatomical images were segmented to obtain a GM probability map. Subsequently, it was coregistered to the first BOLD image in the baseline series and the transformation matrix was applied to the probability map. The coregistered GM probability map was then resliced to the resolution of the BOLD images and a GM binary mask was obtained by thresholding the probability map using a threshold of 0.3.

BOLD processing

BOLD images from all series (baseline, HO, finger tapping and HC) were realigned to the first image in the baseline series and smoothed with a Gaussian filter of 6 mm full-width-at-half-maximum (FWHM).

Task activation analysis

The smoothed images from the finger tapping series were entered into a general linear model (GLM) with two conditions (rest and finger tapping). A regressor was included to model the ASL effect (i.e. the effect of the BOLD image being acquired after ASL control or label acquisitions). A high-pass filter was applied with a cutoff value of 512 s and serial correlations were accounted for using an autoregressive AR(1) model. Activation maps generated by the finger tapping task were assessed with a contrast comparing finger tapping to the rest condition. Obtained activation maps were thresholded by uncorrected p < 0.005 (T>2.72).

Gas stimuli analysis

The smoothed images from the baseline, HO, and HC series were entered into a GLM with three conditions. The model included a regressor to account for the ASL effect and an additional regressor to model the signal linear drift, necessary due to the relatively long duration of the gas stimulus acquisitions. A high-pass filter was applied with a cutoff value of 4096 s and serial correlations were accounted for using an autoregressive AR(1) model. The maps of regression coefficients (beta images) were used to compute the BOLD signal at baseline (BOLD₀) and the signal changes induced by HO (ΔBOLD_HO) and HC (ΔBOLD_HC).

ASL processing

ASL images (both label and control) from all series were realigned to the first ASL image in the baseline series. Label and control images were pair-wise subtracted to generate perfusion-weighted images and smoothed using a Gaussian filter (FWHM=6 mm).

Task activation analysis

Perfusion-weighted images from the finger tapping series were entered into a similar GLM as used for BOLD analysis, excluding the ASL regressor and the autoregressive AR(1) model. Perfusion activation maps were obtained contrasting the finger tapping and rest conditions and thresholded by uncorrected p < 0.005 (T>2.92).

Gas stimuli analysis

Perfusion-weighted images from the other three series (baseline, HO, and HC) were entered into a GLM, as previously described for BOLD, excluding the ASL regressor and the autoregressive model. The obtained beta images were used to compute the perfusion signal at baseline, during HO and HC, and the perfusion changes induced by HO and HC stimuli, which were used to calculate M. In addition, perfusion was quantified in physiologic units as described by Alsop et al.³⁰ with labeling efficiency of 60% due to the background-suppression scheme. Assumed values of T_1,blood were 1660 ms for normoxic conditions³¹ and 1500 ms for HO with FiO₂ ≈ 50%.^16,32,33

Whole-brain Y_v

To compute Y_v, the tilt angle of the SSS with respect to B₀ was measured from the time-of-flight series. Then, the phase difference between the two images separated by ΔTE was computed, and the bias field in the posterior aspect of the brain was fit to a second-order polynomial and corrected as described previously.³⁴ Regions of interest (ROIs) in the SSS and extravascular reference tissue were manually defined, and the phase difference between the vein and background tissue was used to compute whole-brain Y_v from equation (5).

Calculation of [dHb] from P_ETO₂ data

To compare results to the HO calibrated fMRI model proposed by Chiarelli et al.,¹⁰ P_ETO₂ data were converted to [dHb]_v/[dHb]_v0 as described previously. Briefly, the arterial oxygen tension, inferred from P_ETO₂, was converted to arterial oxygen saturation.³⁵ Then, the content of oxygen in arterial (CaO₂) and venous (CvO₂) blood was computed, assuming a constant OEF of 30%. CvO₂ was converted into Y_v for both the baseline and HO conditions. The ratio of (1−Y_v) during HO to that at rest was ultimately used as the input ([dHb]_v/[dHb]_v0) to the P_ETO₂-based calibration model. Calculated [dHb]_v/[dHb]_v0 from P_ETO₂ were also compared to the ratio of (1−Y_v)_HO/(1−Y_v)₀ derived from MR-acquired data.

Quantification of M maps

In all cases, α = 0.18³⁶ and β = 1.5³⁷ were assumed, and four M maps were computed for each subject based on the following calibration approaches:

HO Y_v-based: direct measurement of whole-brain Y_v using baseline and HO stimulus data, with equation (6).

HO P_ETO₂-based: model proposed by Chiarelli et al.¹⁰ using baseline and HO stimulus data with [dHb]_v derived from P_ETO₂ data, using equation (4).

HC Y_v-based: direct measurement of whole-brain Y_v using baseline and HC stimulus data, with equation (6).

HC isometabolism: model proposed by Davis et al.,³ using baseline and HC stimulus data and assuming isometabolism, following equation (3).

In all cases, the CBF/CBF₀ ratio was replaced by the ratio of perfusion-weighted signals. A correction factor was applied to the perfusion signal ratio to account for the shortened T₁ of blood during HO.³⁸ Specifically, the relative perfusion ratio between HO and normoxia conditions was multiplied by 1.15 to account for the change in T₁.

The mean value of M in GM was computed in a mask obtained from the intersection of BOLD images thresholded to a relative intensity of >0.8 (to exclude areas of BOLD signal loss) and the GM mask computed from segmentation of the T₁-MPRAGE. M values were averaged within this mask, excluding outliers, defined a priori. For HC calibration, a voxel M value was considered an outlier if: the perfusion change due to HC was negative or >200%, the BOLD signal change was negative or >15%, or the M value was negative or >40%. For the other two cases, outliers were defined as voxels with perfusion changes >200%, with the other conditions remaining the same.

Relative CMRO₂ response

Relative CMRO₂ (rCMRO₂) maps during finger tapping were computed on a voxel-wise basis as

\begin{matrix} \frac{CMR O_{2}}{CMR O_{2 | 0}} = (\frac{CBF}{CB F_{0}})^{(1 - α / β)} (1 - \frac{\frac{Δ BOLD}{BOL D_{0}}}{M})^{1 / β} \end{matrix}

(7)

where M corresponds to the M map derived from the calibration experiments. Therefore, rCMRO₂ was computed from the same finger tapping task based on the four different calibration results.

Prior to obtaining the mean, voxels were discarded if the relative CMRO₂ was reduced (e.g. rCMRO₂ < 1) or more than doubled (e.g. rCMRO₂ > 2). Average rCMRO₂ in the motor cortex (MC) was computed by averaging values in left and right MC ROIs that were defined by the intersection of areas of activation in the BOLD and ASL finger tapping activation maps.

Statistical analysis

Due to the somewhat limited sample size, a normal distribution cannot be assumed, and therefore non-parametric tests were used for statistical analysis. Wilcoxon signed-rank tests were used to compare the gas stimulus response to baseline for changes in P_ETO₂, P_ETCO₂, and Y_v (or [dHb]_v). Differences between M values and relative CMRO₂ changes obtained from the four calibration models (HO Y_v-based, HO P_ETO₂-based, HC Y_v-based, and HC isometabolism calibration) were evaluated by means of the Friedman test (non-parametric corollary of the repeated measures ANOVA). If the Friedman test was significant, post hoc paired Wilcoxon signed-rank tests were used to test for individual differences. For all tests, p < 0.05 was considered significant. Agreement between the mean GM M and relative CMRO₂ changes derived from the HO Y_v-based, HO P_ETO₂-based, HC Y_v-based, and HC isometabolism calibration models was assessed via Bland–Altman analyses.

Results

Incomplete data were obtained for three subjects: one withdrew due to discomfort during HC, in one subject, the RespirAct system failed to target the desired P_ETO₂ and P_ETCO₂ values, and for the third subject, the ASL labeling plane was incorrectly placed. The remaining 10 subjects are included in subsequent analyses.

Gas manipulations

Figure 2 shows targeted and measured P_ETO₂ and P_ETCO₂, monitored by the RespirAct over the course of the OxBOLD scans in a representative subject. Across all subjects, the mean ( ± standard deviation) of baseline P_ETO₂ was 151 ± 6 mmHg and P_ETCO₂ was 39 ± 5 mmHg. During the HO stimulus period, P_ETO₂ increased to 352 ± 15 mmHg (p < 0.005), with no change in P_ETCO₂ (=39 ± 5 mmHg, p = 0.77). During the HC stimulus, P_ETCO₂ increased to 47 ± 6 mmHg (p < 0.005), with no change in P_ETO₂ (=154 ± 7 mmHg, p = 0.13).

Figure 2.

Targeted (line) and measured (dots) end-tidal oxygen (top) and carbon dioxide (bottom) data in a representative subject. OxBOLD data were acquired during baseline (normoxic/normocapnic air), hyperoxia (+230 mmHg from baseline), and hypercapnia (+8 mmHg from baseline) periods. The finger tapping task was performed while breathing normocapnic/normoxic air between the hyperoxia and hypercapnia periods to make efficient use of scan time. The finger tapping task was 70 s off, then 70 s on, repeated twice. In most subjects, the nominal targeted baseline P_ETO₂ was lower than the measured P_ETO₂, as is shown here.

BOLD, ASL, and Y_v changes during gas stimuli

Figure 3 shows the baseline and gas stimulus images for BOLD, perfusion, and Y_v. BOLD and perfusion data are shown in a representative slice, though both acquisitions have full-brain coverage. Table 1 summarizes the response to the gas stimuli, quantified M for the various models, and relative CMRO₂ responses to the finger tapping task. CBF was not significantly different between baseline and HO (p = 0.28), but as expected, a large increase in blood flow occurred during HC (p < 0.005). More modest ΔBOLD signal changes were observed between baseline and HO than between baseline and HC (p < 0.005).

Figure 3.

Baseline and gas stimulus images for ASL (signal intensity, SI, difference from control and label acquisitions), ΔBOLD, and susceptometry-based oximetry (SBO). Left panel shows anatomic images for reference. Perfusion-weighted signal was relatively unchanged in response to hyperoxia but increased substantially during hypercapnia. The BOLD response was similarly more prominent in response to hypercapnia, though both hyperoxia and hypercapnia stimuli elicited a positive BOLD response. The phase data were converted to venous oxygen saturation through the SBO model. The phase contrast between the superior sagittal sinus (indicated by the arrow) decreases during hyperoxia and hypercapnia, corresponding to an increase in venous oxygen saturation.

Table 1.

Mean ± standard deviation of measurements during baseline and gas stimulus periods, calculated calibration constants, and relative CMRO₂ responses to finger tapping.

	Baseline	Hyperoxia		Hypercapnia
End-tidal
P_ETO₂ (mmHg)	151 ± 6	352 ± 15*		154 ± 7
P_ETCO₂ (mmHg)	39 ± 5	39 ± 5		47 ± 6*
OxBOLD
CBF (mL/min/100 g)	46 ± 10	44 ± 10		67 ± 17*
ΔBOLD/BOLD₀ (%)		2.5 ± 1.0*		3.8 ± 1.9*
Y_v (%HbO₂)	61 ± 8	67 ± 7*		75 ± 10*
		Y_v-based	P_ETO₂-based	Y_v-based	Isometabolic

Gray matter M (%)		11.7 ± 4.5	8.5 ± 2.8	8.9 ± 5.2	9.3 ± 3.8
MC rCMRO₂ (%)		22.0 ± 9.4	20.8 ± 8.2	19.3 ± 9.0	20.5 ± 9.4

P_ETO₂ and P_ETCO₂: the partial pressure of end-tidal oxygen and carbon dioxide, respectively; CBF: cerebral blood flow; Y_v: venous oxygen saturation; M: calibration constant; MC: motor cortex; rCMRO₂: relative cerebral metabolic rate of oxygen in response to finger tapping.

p < 0.005 for difference between stimulus period and baseline.

As measured by SBO, Y_v in the SSS at baseline was 61 ± 8% HbO₂, which increased to 67 ± 7% HbO₂ during HO (p < 0.005), and 75 ± 10% HbO₂ during HC (p < 0.005). The mean tilt angle of the SSS relative to B₀ was 15 ± 10°. For input into the HO P_ETO₂-based calibration model, measured P_ETO₂ was converted to relative [dHb]. Using this approach, [dHb]_v/[dHb]_v0 decreased to 0.90 ± 0.01 during HO relative to the baseline condition. As a direct comparison, the ratio of hyperoxic to baseline (1−Y_v)/(1−Y_v)₀ derived from SBO was 0.85 ± 0.10. The difference between P_ETO₂-derived [dHb]_v changes and the SBO-derived (1−Y_v) response was not significant (p = 0.32).

M maps

Examples of M maps derived from the proposed Y_v-based model from HO and HC stimulus data are compared to M maps based on the standard calibration models in Figure 4. M maps demonstrate expected gray-white matter anatomic contrast for each of the models. The percentage of voxels after outlier exclusion did not differ between calibration methods. Mean % of GM voxels included in the final analysis was 71 ± 18% (HO Y_v-based), 74 ± 15% (HC Y_v-based), 74 ± 18% (HC isometabolism), 72 ± 17% (HO P_ETO₂-based), p = 0.43. Mean M in GM was 11.7 ± 4.5% (HO Y_v-based), 8.9 ± 5.2% (HC Y_v-based), 9.3 ± 3.8% (HC isometabolism), and 8.5 ± 2.8% (HO P_ETO₂-based). The difference between the average M quantified using the various calibration methods was not significant (p = 0.56). Results of the Bland–Altman analysis are shown in Figure 5. Comparing each of the methods, the mean bias of M was no greater than 3.2 (in M, %) with limits of agreement encompassing 0 in all cases.

Figure 4.

M maps showing axial acquisition, and sagittal and coronal reformats in a representative subject. M maps were obtained from hyperoxia (HO) and hypercapnia (HC) gas stimulus data using the proposed Y_v-based model and conventional models. Maps demonstrate relative mutual agreement in this representative subject.

Figure 5.

Bland–Altman analysis plots assessing the agreement of M (top series) and rCMRO₂ (bottom series) between the various calibration models. Overall the mean bias was near zero, and the calculated limits of agreement (±1.96×standard deviation) encompassed zero.

Task-evoked CMRO₂ changes

The finger tapping task activated regions in the motor cortex in both the ASL and BOLD (and thus rCMRO₂) images. Across all subjects, CBF increased by 30.7 ± 9.4% in the activated regions, whereas the BOLD-weighted signal increased by 0.9 ± 0.2%. Based on voxel-wise M maps from each calibration model, CMRO₂ in the motor cortex increased by 22.0 ± 9.4% (HO Y_v-based model), 19.3 ± 9.0% (HC Y_v-based model), 20.5 ± 9.4% (HC isometabolism model), and 20.8 ± 8.2% (HO P_ETO₂-based model). The difference of rCMRO₂ among the four models was not significant (p = 0.25). Relative CMRO₂ changes in a representative subject are shown in Figure 6, demonstrating both localization and magnitude of the CMRO₂ response. Bland–Altman results are shown in Figure 5. As in M, the mean bias of rCMRO₂ was close to 0 for all comparisons.

Figure 6.

Relative CMRO₂ change in response to a finger tapping task in a representative subject. Focal CMRO₂ increases are observed in the motor cortex. While these data are derived from the same task-based BOLD and ASL data, the M maps used to convert the relative BOLD and CBF changes differ.

Discussion

This work demonstrates the utility of OxBOLD combined with the Y_v-based model for calibrated fMRI. Using the Y_v-based model, average M values and maps were similar to those derived from accepted models for both HO and HC stimuli. The proposed method directly measures whole-brain Y_v, rather than relying on conversion of end-tidal O₂, and does not mandate an assumption on baseline OEF or of isometabolism as is required for the models described by Chiarelli et al.¹⁰ and Davis et al.³ The mean M values in GM were in mutual agreement over the four calibration experiments. Derived from identical task response data, rCMRO₂ changes in the motor cortex calculated based on Y_v-based M maps agree well with the relative response obtained from existing models' M maps as well.

The critical assumption made for Y_v-based calibration is that gas stimulus-induced changes in Y_v across the brain are spatially uniform. Note that this does not require that baseline OEF is constant (though there is evidence from MRI³⁹ and PET⁴⁰ studies that OEF is quite uniform over the brain). As long as OEF changes are spatially uniform, the model would apply irrespective of metabolic changes (or lack thereof). Thus, in addition to the calibration stimuli explored in this study, Y_v-based calibration could theoretically be applied in the context of other global vasoactive stimuli that are not necessarily isometabolic. One could, therefore, explore alternative stimuli such as breath-hold,⁴¹ hypoxia,⁴² caffeine,⁴³ or possibly anesthesia.⁴⁴ This is unlike the HO P_ETO₂-based and HC isometabolism model, where an isometabolic response to the calibration stimulus is inherently assumed. Recently, there have been conflicting reports regarding the metabolic response to these stimulus gases, where some report isometabolism to HC^45,46 and HO,⁴⁷ and others have found CMRO₂ decreases in response to even mild HO or HC.¹⁵

In consideration of the inherent assumption underlying the Y_v-based model, studies investigating BOLD and CBF-based cerebrovascular reactivity (CVR) indicate non-uniform CVR throughout the cortex in response to HC, with smaller changes in WM compared to GM.⁴⁸ If HC is indeed isometabolic, then GM/WM CVR differences would imply that WM and GM OEF changes are also not spatially uniform, thus contradicting the necessary assumption for Y_v-based calibration. However, this may not impart a measurable error in the data. Assuming that conservatively, CBF in WM is half that of GM,^49,50 and that the volume of WM is again about half that of GM,⁵¹ then blood flow from GM contributes to 80% of the total blood flow in the SSS under resting conditions. CBF-based CVR studies have reported a 1.9%/mmHg CBF change in GM compared to 1.1%/mmHg change in WM.⁴⁸ During a + 8 mmHg HC challenge, flow will therefore increase in both GM and, to a somewhat lesser extent, WM. If we assume that baseline OEF is uniformly 30.0% and HC is isometabolic, the CVR-based flow changes imply OEF in GM is reduced to 26.0%, whereas OEF in WM is reduced to 27.6% during HC (calculated via Fick's principle). Given the GM/WM blood flow contribution to blood in the SSS (which would favor GM slightly more during HC), the measured OEF in the SSS would be 26.3%. As the focus for this calibrated fMRI study is on GM, the WM/GM CVR differences would thus contribute an imperceptibly small offset in the SBO data (Δϕ < 0.005 radians, (equation (5)).

Prior reports imply that HO-induced flow changes are small and generally spatially uniform,¹⁶ thus suggesting that the assumption inherent to the Y_v-based model would be valid in the context of a HO stimulus. In fact, more stringent assumptions are placed on the P_ETO₂-based model¹⁰: OEF is assumed to be invariant to HO; and [dHb]_v changes are assumed as constant throughout the entire body since the expired air is equilibrated with the systemic circulation. It also should be noted that changes in [dHb]_v in response to HC cannot be inferred from P_ETO₂.

Comparing HO calibration models, we note that in the Y_v-based model, we did not include the CBF correction term that is employed in the P_ETO₂-based model proposed by Chiarelli et al.¹⁰ By including the CBF correction term, the model does not reduce to the unmodified calibrated fMRI equation (equation (2)) proposed by Davis et al.³ unless isometabolism is specifically assumed or CBF changes do not occur. As with HC, there is some debate about the physiologic response to HO, with some reports showing no change in CMRO₂,⁵² while other work suggests depression of CMRO₂.¹⁵ Furthermore, accurate quantification of CBF changes in response to HO requires knowledge of arterial blood T₁, which shortens due to the presence of plasma-dissolved oxygen.³⁸

Despite the differences in the assumptions and calibration equations, the mean M and relative CMRO₂ response were in mutual agreement between the methods. There is also reasonable agreement between the mean M reported herein and values reported previously. As described in the introduction, M represents the maximum BOLD signal change with full washout of deoxyhemoglobin, and is related to vessel geometry and size, baseline OEF, field strength, and TE. As summarized in Gauthier et al.⁵³ and Lajoie, et al.,⁵⁴ the mean M from prior reports, here corrected for the OxBOLD TE=29 ms, ranges from 5.1%⁵⁵ to 11.7%.⁵⁶ Thus, the mean M reported herein is on the higher end, especially for HO Y_v-based calibration data, but within the range of previous reports. The relative CMRO₂ response of ∼19–22% to finger tapping is similar to findings by Tak et al.,⁵⁷ who reported an 18% CMRO₂ increase in the motor cortex. Our present results are, however, somewhat higher than a prior report from our group,²⁶ which found rCMRO₂ ∼13%. The discrepancy herein could be attributed to the removal of outliers, where in the present study only voxels with a positive CMRO₂ response were ultimately included in the motor cortex ROI. Another possible reason for discrepancy between our work and prior studies is the chosen values for exponents α and β. As detailed by Griffeth and Buxton,⁵⁸ the accuracy and precision of M and rCMRO₂ results depend on appropriate selection of these parameters. More recently, Shu et al.⁵⁹ proposed mapping β, the factor relating intra- and extra-vascular BOLD signal contributions, rather using a single value across the brain. However, specific impacts of varying the constants in the calibration equations were not evaluated herein.

The OxBOLD sequence follows the dual-acquisition approach for ASL and BOLD, which provides improved temporal signal-to-noise ratio of CBF due to optimized background suppression, while maintaining BOLD temporal signal-to-noise ratio.²⁶ Although the k_x–k_y readout trajectory and spatial coverage are identical, the difference in encoding strategy (3D for ASL versus 2D for BOLD) leads to distinct distortion and blurring artifacts between the two sets of images. The distortions are most pronounced at air-tissue interfaces, such as the nasal sinuses and auditory canals, but notably the motor cortex is unimpacted. To generate M maps and compute rCMRO₂, the BOLD and ASL data are combined on a voxel-wise basis. Therefore, these distortions may impact the overall accuracy of the results, but would influence all calibration methods similarly.

The Y_v-based model could theoretically be applied to independently acquired BOLD, ASL, and Y_v data, as preliminarily explored for HO calibration by Driver et al.¹¹ However, sequential acquisition would be less time-efficient, and may be prone to physiologic drift over the course of the stimulus. T₂-based oximetry may alternatively be used to quantify whole-brain Y_v in the SSS^15,19 or other veins such as the straight sinus.⁶⁰ However, if included in an interleaved sequence, the overall TR would be prolonged to accommodate multiple T₂-preparation and readout periods. As veins become closer to perpendicular to B₀, SBO cannot be used to quantify Y_v since the induced field outside the vessel, which is ignored in equation (5), becomes increasingly non-uniform. However, simulations show only modest error is introduced for vessels that deviate <30° from B₀,⁶¹ and the tilt angle of SSS is generally well within this bound.

Ideally, global (whole-brain or whole-body) measures of [dHb]_v would be replaced by voxel-wise maps of OEF, which would additionally permit quantification of both resting and stimulus-evoked CMRO₂ in physiologic units. In a method initially proposed by Bulte et al.,⁶ a hypercapnic stimulus was used to determine M, and subsequently combined with BOLD and ASL signal changes in response to HO to quantify baseline OEF on a voxel-wise basis. Alternatively, Gauthier et al.⁵ proposed use of a generalized calibration model¹³ in which fractional changes in BOLD, CBF, and arterial oxygen content are used to estimate M as a function of baseline OEF. By incorporating both HO and HC data, the voxel-wise solution for M and OEF is obtained as the intersection of the HO- and HC-derived M versus resting OEF curves. Furthermore, evaluating the response to carbogen (a hypercapnic/hyperoxic stimulus), the authors showed that the intersection of all three curves occurred coincidentally. As is the case for M mapping, use of the Y_v-based model in the context of dual-gas calibration may help to alleviate some inherent assumptions.

Other studies have investigated gas-free methods to calibrate the BOLD signal or otherwise quantify OEF at rest. Most of these methods relate the reversible component of the effective transverse relaxation rate (R₂′) to local [dHb].⁶² For example, Blockley et al.^63,64 investigated the use of a gas-free refocusing-based calibration in which R₂- and R₂′-sensitive imaging were used to determine a calibration constant related to M. Berman et al.⁶⁵ recently extended this gas-free calibration approach to additionally correct for the effect of variation in vessel size. Alternatively, by modeling the capillary network as randomly oriented cylinders, R₂′ can be directly related to Y_v by means of what is generally referred to as ‘quantitative BOLD’ (qBOLD).³⁹ Combining voxel-wise estimates of OEF from qBOLD with CBF maps from ASL permits quantification of local CMRO₂ via Fick's principle. However, to date, qBOLD sequences are time prohibitive if whole-brain coverage is desired,⁶⁶ and thus cannot quantify changes in CMRO₂ in response to a stimulus. Another approach to quantify CMRO₂ focuses on the parenchymal susceptibility changes in response to a global stimulus, causing, for instance, vasoconstriction as achieved by caffeine administration or hyperventilation. In this manner, Zhang et al.^67,68 obtained maps of CMRO₂ using constituent susceptibilities (obtained from quantitative susceptibility mapping⁶⁹) and voxel-wise CBF. While this approach is potentially powerful, the small change in tissue susceptibility in response to a challenge (∼+0.005 ppm for caffeine) is difficult to precisely measure. Furthermore, a baseline and stimulus measurement are required to generate a CMRO₂ map, thus, like qBOLD, the technique is limited to probing CMRO₂ in the steady state.

The results reported herein show that Y_v-based calibration can be applied in the context of either HO or HC stimuli. In the healthy subjects recruited for this study, HO was better tolerated than HC, and therefore HO may be preferred for future clinical adaptation of calibrated fMRI. Furthermore, there is widespread availability of oxygen in most hospital settings, whereas HC, requiring controlled delivery of CO₂ mixed in room air, maybe more difficult to broadly disseminate. Translation of Y_v-based calibration to clinical patient populations will require careful consideration of the critical assumption of uniform changes in Y_v in response to the stimulus condition. This assumption would likely hold true for the study rCMRO₂ in patients with degenerative brain disorders, but maybe less appropriate in the context of local perturbations of CBF (e.g. ischemic stroke).

Conclusions

MR-measured whole-brain Y_v can be incorporated into calibrated fMRI models that relate BOLD, CBF, and CMRO₂ changes in response to a stimulus. The Y_v-based model can be equivalently applied to HO or HC gas stimuli, and it relaxes the underlying assumption of isometabolism in response to the gas stimulus as all of the input parameters are directly measured. Realization of the Y_v-based model was achieved with the OxBOLD sequence, providing concurrent measurement of CBF, BOLD-weighted signal, and whole-brain Y_v. Overall agreement was found between traditional calibration models and the Y_v-based calibration both in terms of the calibration constant, M, and the relative CMRO₂ response to finger tapping.

Footnotes

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institutes of Health grants R21 EB022687, T32 EB020087, and P41 EB015893.

Acknowledgements

The authors wish to thank Kelly Sexton, Morgan Kelly and Kathleen Thomas for support with subject recruitment, and Norman Butler, Patricia O'Donnell, Doris Cain and Jacqui Meeks for their assistance with the MRI scans.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors' contributions

EKE, MAFS, AERS, ZBR, and FWW conceived and designed this study. EKE, MAFS, AERS, HL, MV, and ZBR contributed to the pulse sequence design, implementation, and reconstruction and analysis methods. EKE, MAFS, AERS, and HL collected and analyzed data. EKE, MAFS, AERS, HL, and FWW contributed to data interpretation. EKE and MAFS drafted the manuscript. AERS, HL, MV, ZBR, JAD, and FWW critically revised the manuscript. All authors approved the final version of the manuscript for publication.

References

Ogawa

Lee

Kay

, et al. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci U S A 1990; 87: 9868–9872.

Kim

S-G

Uğurbil

. Comparison of blood oxygenation and cerebral blood flow effects in fMRI: estimation of relative oxygen consumption change. Magn Reson Med 1997; 38: 59–65.

Davis

Kwong

Weisskoff

, et al. Calibrated functional MRI: mapping the dynamics of oxidative metabolism. Proc Natl Acad Sci U S A 1998; 95: 1834–1839.

Hoge

Atkinson

Gill

, et al. Investigation of BOLD signal dependence on cerebral blood flow and oxygen consumption: the deoxyhemoglobin dilution model. Magn Reson Med 1999; 42: 849–863.

Gauthier

Desjardins-Crépeau

Madjar

, et al. Absolute quantification of resting oxygen metabolism and metabolic reactivity during functional activation using QUO2 MRI. Neuroimage 2012; 63: 1353–1363.

Bulte

Kelly

Germuska

, et al. Quantitative measurement of cerebral physiology using respiratory-calibrated MRI. Neuroimage 2012; 60: 582–591.

Wise

Harris

Stone

, et al. Measurement of OEF and absolute CMRO₂: MRI-based methods using interleaved and combined hypercapnia and hyperoxia. Neuroimage 2013; 83: 135–147.

Kim

S-G

Rostrup

Larsson

, et al. Determination of relative CMRO₂ from CBF and BOLD changes: significant increase of oxygen consumption rate during visual stimulation. Magn Reson Med 1999; 41: 1152–1161.

Leontiev

Buxton

. Reproducibility of BOLD, perfusion, and CMRO₂ measurements with calibrated-BOLD fMRI. Neuroimage 2007; 35: 175–184.

10.

Chiarelli

Bulte

Wise

, et al. A calibration method for quantitative BOLD fMRI based on hyperoxia. Neuroimage 2007; 37: 808–820.

11.

Driver

Hall

Wharton

, et al. Calibrated BOLD using direct measurement of changes in venous oxygenation. Neuroimage 2012; 63: 1178–1187.

12.

Krieger

Ivanov

Huber

, et al. Using carbogen for calibrated fMRI at 7Tesla: comparison of direct and modelled estimation of the M parameter. Neuroimage 2014; 84: 605–614.

13.

Gauthier

Hoge

. A generalized procedure for calibrated MRI incorporating hyperoxia and hypercapnia. Hum Brain Mapp 2012; 34: 1053–1069.

14.

Battisti-Charbonney

Fisher

Duffin

. The cerebrovascular response to carbon dioxide in humans. J Physiol 2011; 589: 3039–3048.

15.

Liu

Pascual

, et al. Effect of hypoxia and hyperoxia on cerebral blood flow #blood |oxygenation, and oxidative metabolism. J Cereb Blood Flow Metab 2012; 32: 1909–1918.

16.

Bulte

Chiarelli

Wise

, et al. Cerebral perfusion response to hyperoxia. J Cereb Blood Flow Metab 2006; 27: 69–75.

17.

Losert

Peller

Schneider

, et al. Oxygen-enhanced MRI of the brain. Magn Reson Med 2002; 48: 271–277.

18.

Prisman

Slessarev

Han

, et al. Comparison of the effects of independently-controlled end-tidal PCO₂ and PO₂ on blood oxygen level–dependent (BOLD) MRI. J Magn Reson Imaging 2007; 27: 185–191.

19.

. Quantitative evaluation of oxygenation in venous vessels using T₂-relaxation-under-spin-tagging MRI. Magn Reson Med 2008; 60: 357–363.

20.

Fernández-Seara

Techawiboonwong

Detre

, et al. MR susceptometry for measuring global brain oxygen extraction. Magn Reson Med 2006; 55: 967–973.

21.

Jain

Langham

Wehrli

. MRI estimation of global brain oxygen consumption rate. J Cereb Blood Flow Metab 2010; 30: 1598–1607.

22.

Grubb

Jr Raichle

Stroke

. The effects of changes in PaCO₂ cerebral blood volume #blood |flow, and vascular mean transit time. Stroke 1974; 5: 630–639.

23.

Spees

Yablonskiy

Oswood

, et al. Water proton MR properties of human blood at 1.5 Tesla: magnetic susceptibility, T₁, T₂, T₂* and non-Lorentzian signal behavior. Magn Reson Med 2001; 45: 533–542.

24.

Jain

Abdulmalik

Propert

, et al. Investigating the magnetic susceptibility properties of fresh human blood for noninvasive oxygen saturation quantification. Magn Reson Med 2011; 68: 863–867.

25.

Vidorreta

Wang

Chang

, et al. Whole-brain background-suppressed pCASL MRI with 1D-accelerated 3D RARE Stack-Of-Spirals readout. PLoS One 2017; 12: e0183762–16.

26.

Fernández-Seara

Rodgers

Englund

, et al. Calibrated bold fMRI with an optimized ASL-BOLD dual-acquisition sequence. Neuroimage 2016; 142: 474–482.

27.

Maleki

Dai

Alsop

. Optimization of background suppression for arterial spin labeling perfusion imaging. Magn Reson Mater Phys 2011; 25: 127–133.

28.

van Vaals

Brummer

Dixon

, et al. ‘Keyhole’ method for accelerating imaging of contrast agent uptake. J Magn Reson Imaging 1993; 3: 671–675.

29.

Slessarev

Han

Mardimae

, et al. Prospective targeting and control of end-tidal CO₂ and O₂ concentrations. J Physiol 2007; 581: 1207–1219.

30.

Alsop

Detre

Golay

, et al. Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: a consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia. Magn Reson Med 2014; 73: 102–116.

31.

Clingman

Golay

, et al. Determining the longitudinal relaxation time (T₁) of blood at 3.0 Tesla. Magn Reson Med 2004; 52: 679–682.

32.

Siero

JCW

Strother

Faraco

, et al. In vivo quantification of hyperoxic arterial blood water T₁. NMR Biomed 2015; 28: 1518–1525.

33.

Pilkinton

Hiraki

Detre

, et al. Absolute cerebral blood flow quantification with pulsed arterial spin labeling during hyperoxia corrected with the simultaneous measurement of the longitudinal relaxation time of arterial blood. Magn Reson Med 2011; 67: 1556–1565.

34.

Langham

Magland

Floyd

, et al. Retrospective correction for induced magnetic field inhomogeneity in measurements of large-vessel hemoglobin oxygen saturation by MR susceptometry. Magn Reson Med 2009; 61: 626–633.

35.

Severinghaus

. Simple, accurate equations for human blood 02 dissociation computations. J Appl Physiol 1979; 46: 599–602.

36.

Germuska

Wise

. Calibrated fMRI for mapping absolute CMRO₂: practicalities and prospects. Neuroimage 2019; 187: 145–153.

37.

Boxerman

Hamberg

Rosen

, et al. MR contrast due to intravascular magnetic susceptibility perturbations. Magn Reson Med 1995; 34: 555–566.

38.

Tadamura

Hatabu

, et al. Effect of oxygen inhalation on relaxation times in various tissues. J Magn Reson Imaging 1997; 7: 200–225.

39.

Yablonskiy

. Quantitative BOLD: mapping of human cerebral deoxygenated blood volume and oxygen extraction fraction: default state. Magn Reson Med 2006; 57: 115–126.

40.

Raichle

MacLeod

Snyder

, et al. A default mode of brain function. Proc Natl Acad Sci U S A 2001; 98: 676–682.

41.

Rodgers

Jain

Englund

, et al. High temporal resolution MRI quantification of global cerebral metabolic rate of oxygen consumption in response to apneic challenge. J Cereb Blood Flow Metab 2013; 33: 1514–1522.

42.

Vestergaard

Lindberg

Aachmann-Andersen

, et al. Acute hypoxia increases the cerebral metabolic rate – a magnetic resonance imaging study. J Cereb Blood Flow Metab 2016; 36: 1046–1058.

43.

Griffeth

VEM

Perthen

Buxton

. Prospects for quantitative fMRI: investigating the effects of caffeine on baseline oxygen metabolism and the response to a visual stimulus in humans. Neuroimage 2011; 57: 809–816.

44.

Oshima

Karasawa

Satoh

. Effects of propofol on cerebral blood flow and the metabolic rate of oxygen in humans. Acta Anaesthesiol Scand 2002; 46: 831–835.

45.

Chen

Pike

. Global cerebral oxidative metabolism during hypercapnia and hypocapnia in humans: implications for BOLD fMRI. J Cereb Blood Flow Metab 2010; 30: 1094–1099.

46.

Jain

Langham

Floyd

, et al. Rapid magnetic resonance measurement of global cerebral metabolic rate of oxygen consumption in humans during rest and hypercapnia. J Cereb Blood Flow Metab 2011; 31: 1504–1512.

47.

Diringer

Aiyagari

Zazulia

, et al. Effect of hyperoxia on cerebral metabolic rate for oxygen measured using positron emission tomography in patients with acute severe head injury. J Neruosurg 2007; 106: 526–529.

48.

Halani

Kwinta

Golestani

, et al. Comparing cerebrovascular reactivity measured using BOLD and cerebral blood flow MRI: the effect of basal vascular tension on vasodilatory and vasoconstrictive reactivity. Neuroimage 2015; 110: 110–123.

49.

Pantano

Baron

J-C

Lebrun-Grandie

, et al. Regional cerebral blood flow and oxygen consumption in human aging. Stroke 1984; 15: 635–641.

50.

W-C

Lin

S-C

Wang

, et al. Measurement of cerebral white matter perfusion using pseudocontinuous arterial spin labeling 3T magnetic resonance imaging – an experimental and theoretical investigation of feasibility. PLoS One 2013; 8: e82679.

51.

Luders

Steinmetz

Jancke

. Brain size and grey matter volume in the healthy human brain. Cognitive Neurosci Neuropsychol 2002; 13: 2371–2374.

52.

Sicard

Duong

. Effects of hypoxia, hyperoxia, and hypercapnia on baseline and stimulus-evoked BOLD, CBF, and CMRO₂ in spontaneously breathing animals. Neuroimage 2005; 25: 850–858.

53.

Gauthier

Madjar

Tancredi

, et al. Elimination of visually evoked BOLD responses during carbogen inhalation: implications for calibrated MRI. Neuroimage 2011; 54: 1001–1011.

54.

Lajoie

Tancredi

Hoge

. Regional reproducibility of BOLD calibration parameter M, OEF and resting-state CMRO₂ measurements with QUO₂ MRI. PLoS One 2016; 11: e0163071.

55.

Bulte

Drescher

Jezzard

. Comparison of hypercapnia-based calibration techniques for measurement of cerebral oxygen metabolism with MRI. Magn Reson Med 2009; 61: 391–398.

56.

Perthen

Lansing

Liau

, et al. Caffeine-induced uncoupling of cerebral blood flow and oxygen metabolism: a calibrated BOLD fMRI study. Neuroimage 2008; 40: 237–247.

57.

Tak

Jang

Lee

, et al. Quantification of CMRO₂ without hypercapnia using simultaneous near-infrared spectroscopy and fMRI measurements. Phys Med Biol 2010; 55: 3249–3269.

58.

Griffeth

VEM

Buxton

. A theoretical framework for estimating cerebral oxygen metabolism changes using the calibrated-BOLD method: modeling the effects of blood volume distribution, hematocrit, oxygen extraction fraction, and tissue signal properties on the BOLD signal. Neuroimage 2011; 58: 198–212.

59.

Shu

Sanganahalli

Coman

, et al. Quantitative β mapping for calibrated fMRI. Neuroimage 2016; 126: 219–228.

60.

Jain

Magland

Langham

, et al. High temporal resolution in vivo blood oximetry via projection-based T2 measurement. Magn Reson Med 2012; 70: 785–790.

61.

Langham

Epstein

, et al. Accuracy of the cylinder approximation for susceptometric measurement of intravascular oxygen saturation. Magn Reson Med 2011; 67: 808–813.

62.

Yablonskiy

Haacke EM . Theory of Nmr signal behavior in magnetically inhomogeneous tissues – the static dephasing regime. Magn Reson Med 1994; 32: 749–763.

63.

Blockley

Griffeth

VEM

Buxton

. A general analysis of calibrated BOLD methodology for measuring CMRO₂ responses: comparison of a new approach with existing methods. Neuroimage 2012; 60: 279–289.

64.

Blockley

Griffeth

VEM

Simon

, et al. Calibrating the BOLD response without administering gases: comparison of hypercapnia calibration with calibration using an asymmetric spin echo. Neuroimage 2015; 104: 423–429.

65.

Berman

AJL

Mazerolle

MacDonald

, et al. Gas-free calibrated fMRI with a correction for vessel-size sensitivity. Neuroimage 2018; 169: 176–188.

66.

Lee

Englund

Wehrli

. Interleaved quantitative BOLD: combining extravascular R₂′- and intravascular R₂-measurements for estimation of deoxygenated blood volume and hemoglobin oxygen saturation. Neuroimage 2018; 174: 420–431.

67.

Zhang

Liu

Gupta

, et al. Quantitative mapping of cerebral metabolic rate of oxygen (CMRO₂) using quantitative susceptibility mapping (QSM). Magn Reson Med 2014; 74: 945–952.

68.

Zhang

Zhou

Nguyen

, et al. Cerebral metabolic rate of oxygen (CMRO₂) mapping with hyperventilation challenge using quantitative susceptibility mapping (QSM). Magn Reson Med 2016; 77: 1762–1773.

69.

Wu B and Liu

. Quantitative susceptibility mapping of human brain reflects spatial variation in tissue composition. Neuroimage 2011; 55: 1645–1656.

Calibrated fMRI for dynamic mapping of CMRO 2 responses using MR-based measurements of whole-brain venous oxygen saturation

Abstract

Keywords

Introduction

Theory

Materials and methods

Image acquisition

Gas manipulations

Motor stimulus

Data processing and analysis

BOLD processing

Task activation analysis

Gas stimuli analysis

ASL processing

Task activation analysis

Gas stimuli analysis

Whole-brain Yv

Calculation of [dHb] from PETO2 data

Quantification of M maps

Relative CMRO2 response

Statistical analysis

Results

Gas manipulations

BOLD, ASL, and Yv changes during gas stimuli

M maps

Task-evoked CMRO2 changes

Discussion

Conclusions

Footnotes

Funding

Acknowledgements

Declaration of conflicting interests

Authors' contributions

References

Whole-brain Y_v

Calculation of [dHb] from P_ETO₂ data

Relative CMRO₂ response

BOLD, ASL, and Y_v changes during gas stimuli

Task-evoked CMRO₂ changes