Synopsis

Keywords: Quantitative Imaging, Metabolism

In the original qBOLD, it is challenging to separate deoxyhemoglobin’s contribution to R₂' from other sources modulating the voxel signal. Further, extracting DBV and Y_v from measured R₂' is a nontrivial task. It was recently shown that the constrained qBOLD method was able to properly separate the several confounding factors, yielding the expected contrast for both Y_v and DBV maps across the entire brain, and, together with a separate measurement of CBF, leading to whole-brain 3D CMRO₂ maps within physiologically plausible ranges. Here, we validated the new 3D qBOLD method with respect to repeatability and hypercapnic gas breathing challenges.

Introduction

Quantitative BOLD (qBOLD)^1,2 allows evaluation of oxidative metabolism of the brain in the resting-state by quantifying deoxygenated blood volume (DBV) and hemoglobin oxygen saturation level of venous blood (Yv), based on measurements of RF-reversible transverse relaxation rate constant R₂′. However, it is challenging in the original qBOLD to separate deoxyhemoglobin’s contribution to R₂′ from other sources modulating the voxel signal, for instance, R₂, R₂′ from non-heme iron (R_2,nh′), and macroscopic magnetic field inhomogeneities (ΔB₀). Furthermore, the qBOLD model is highly sensitive to noise, resulting in unstable estimation of DBV and Y_v, even with an accurate measurement of the heme-originated R₂′ (R_2,h′)³. To address these issues, we had previously introduced a prior-constrained qBOLD method⁴ in which a plurality of preliminary parametric maps (R₂, R₂′, ΔB₀, voxel susceptibility (Δχ), and cerebral venous blood volume (CBV_v)) are constrained in the qBOLD model. In that work it was shown that the method was able to properly separate the above-mentioned confounders, yielding the expected contrast for both Y_v and DBV maps across the entire brain_⁵, and, together with a separate measurement of cerebral blood flow (CBF), leading to whole-brain 3D cerebral metabolic rate of oxygen (CMRO₂) maps within physiologically plausible ranges⁴. The purpose of the present work was to validate the new 3D qBOLD method with respect to repeatability and hypercapnic gas breathing challenges.

Methods

Summary of constrained qBOLD: The qBOLD protocol in this study consists of two in-house developed pulse sequences: alternating unbalanced SSFP-FID and SSFP-ECHO (AUSFIDE)⁵ (Fig. 1a) and velocity-selective venous-spin-labeling (VS-VSL)⁶ (Fig. 1b), yielding voxelwise maps of 3D R₂, R₂′, ΔB₀, Δχ (Figs. 1c, 1d) and a volumetric map of CBV_v (Fig. 1e), respectively. In the qBOLD processing step, the AUSIFDE-derived parameters serve as prior information in solving the following constrained optimization problem:
$$\arg min_{\mathbf{\Theta}}\sum_{TE}|\mathbf{y}(TE)-\Xi (\mathbf{\Theta},TE))|^{2}+w|\Delta \chi -\Psi (\mathbf{\Theta})|^{2}+p|{{R_{2}}'}-\Upsilon (\mathbf{\Theta})|^{2}$$ [1]
where Θ={Y_v,DBV,R₂′,χ_nb} is the set of unknown parameters (χ_nb: non-blood susceptibility). Definitions for the models, Ξ, Ψ, and Γ, along with sub-parameters therein are provided in Reference 4. In solving Eq. (1), VS-VSL-derived CBV_v maps were used to initialize and guide the solution of DBV. The regularization parameters w and p were empirically determined.
Given the estimates of Y_v across the entire brain, oxygen extraction fraction (OEF) was computed by OEF = 1-Y_v/Y_a with Y_a = 0.98 (arterial blood oxygen saturation) assumed. Finally, 3D pseudo-continuous arterial spin labeling (pCASL) with stack-of-spirals readout⁷ was employed for CBF mapping, leading to the derivation of CMRO₂ by: CMRO₂=C_a·CBF·Y_a·OEF where C_a is the oxygen carrying capacity of arterial blood, obtained by scaling a hematocrit level.
Experiments & analyses: All experiments were performed at 3T (Siemens Prisma). To evaluate the present method’s reproducibility, data were collected twice in seven healthy volunteers using the three pulse sequences (AUSFIDE, VS-VSL, and pCASL). In all study subjects, the second scans took place within two weeks apart from the first one. Additionally, high-resolution T₁-weighted images were also acquired for brain segmentation. See Reference 4 for imaging parameters and data processing details. Both DBV and OEF maps in a representative subject were presented for the two scans. Furthermore, the interscan agreement of the two parameters in gray-matter (GM) was evaluated using box plots and paired t-tests. The same protocol was also applied in 11 healthy participants under baseline and hypercapnic states, respectively, for further validation of the technique. Here, the RespirAct system (Thornhill Research) was used to target +8 mmHg of end-tidal partial pressure of CO₂ from baseline (equivalent to about 6% inspired CO₂). Following data processing, whole-brain 3D maps of CBF, DBV, OEF, and CMRO₂ were obtained in a representative subject at baseline and hypercapnia. Thereafter, GM averages of the four parameters across the 11 study subjects were compared between the two states using paired t-tests.

Results

Figure 2 displays test/retest whole-brain 3D maps of DBV and OEF in the three orthogonal planes in a study subject. In both scans, DBV images highlight the expected gray/white matter contrast while OEF values are largely invariable across the entire brain, being consistent with the results in our previous study⁴. Figure 3 shows box plots of GM averages of DBV and OEF in the seven subjects. Paired t-test suggests that the test-retest difference in both DBV and OEF quantifications is not statistically significant.
Figure 4 shows parametric maps of CBF, DBV, OEF and CMRO₂ reformatted into the three orthogonal planes, obtained in a subject breathing normoxic/normocapnic and hypercapnic gas mixtures. Box plots of GM averages of the four parameters in the two states are shown in Fig. 5. Statistical comparisons reveal that the hypercapnia-induced increases of both CBF and DBV and decrease of OEF were all significant, whereas the CMRO₂ changes due to the hypercapnic stimulus were nonsignificant.

Discussion and Conclusions

We evaluated the performance of the constrained qBOLD based oximetric technique by examining its repeatability and detectability on physiologic changes resulting from external stimuli. Results suggest that the method yields the expected contrast across the brain regions (GM vs. WM), across the measurement times (test vs. retest), and further across the brain states (baseline vs. hypercapnia), albeit so far in healthy subjects. Clinical validation in patients with neurovascular disease, for example, arterial steno-occlusive disease is forthcoming.

Acknowledgements

NIH grants P41-EB029460 and R21-EB031364, and NRF Korea grant 2021R1F1A1045621