0100

Whole-Brain 3D Quantitative BOLD Mapping With Preliminary Estimation of R₂' and Venous Blood Volume

Hyunyeol Lee¹ and Felix W Wehrli¹
¹Radiology, University of Pennsylvania, Philadelphia, PA, United States

Synopsis

Quantitative BOLD (qBOLD) seeks to quantify voxel-wise deoxygenated blood volume (DBV) and venous blood oxygen saturation (Y_v) based on R₂′-sensitive signal acquisitions. A complication in qBOLD is the separation of signal contributions from R₂, and R₂′ from heme and non-heme iron sources, particularly in the deep brain structures. Here, we develop a new 3D qBOLD mapping method by tackling the confounding factors based on preliminary estimates of R₂, R₂’, and voxel susceptibility, along with cerebral venous blood volume. Results suggest feasibility of the proposed, prior-based qBOLD method for 3D mapping of DBV and Y_v across the entire brain.

Introduction

Quantitative BOLD (qBOLD)^1,2 seeks to quantify voxel-wise deoxygenated blood volume (DBV) and venous blood oxygen saturation (Y_v) based on R₂'-sensitive signal acquisitions. However, in the original qBOLD method, it is challenging to separate the two parameters from the estimated R₂' because of their mutual coupling in the model³. To address this issue, we had previously proposed to combine extravascular R₂' and intravascular R₂ mapping approaches for prior information based qBOLD processing⁴. While the method substantially improved stability of the parameter quantification, it only allows 2D single-slice quantification, thus limiting its utility in clinical practice. A further complication in the qBOLD method is the separation of signal contributions from R₂, and R₂' from heme and non-heme iron sources, particularly in the deep brain structures (e.g., putamen, globus pallidus). Thus, the objective of this work was to develop a new 3D MRI oximetry technique that enables robust qBOLD parameter mapping across the entire brain. To this end, we implemented a 3D qBOLD processing pipeline, which takes into account the confounding factors by making use of preliminary estimates of R₂, R₂', and voxel susceptibility (χ) obtained using our recently reported technique, referred to as alternating unbalanced SSFP-FID and SSFP-ECHO (AUSFIDE)⁵, and additionally measured venous cerebral blood volume (CBV_v) via a velocity-selective venous-spin-labeling (VS-VSL) method⁶.

Methods

Pulse sequence & problem definition: The AUSFIDE pulse sequence⁵ permits rapid 3D encoding of both R₂ and R₂' by collecting multiple SSFP-FID and SSFP-ECHO signals alternately over sequence repetitions. Furthermore, the method’s inherent sensitivity to large-scale induced magnetic field offsets (ΔB₀) was tackled by means of 3D z-shimming and voxel-spread-function (VSF)-based correction^7,8. Figure 1 shows the proposed qBOLD processing pipeline, in which the AUSFIDE-derived parameters, R₂, R₂', and χ, are employed to serve as preliminary estimates in solving the following problem:
$$\underset{\Theta}{\arg\min}\sum_{t}\lvert y(t)-\Xi(\Theta,t)\rvert^2+w\lvert \chi-\Psi(\Theta) \rvert^2+p\lvert R_2'-\Gamma(\Theta) \rvert^2 (1) $$
Here, y(t) is the acquired multi-echo AUSFIDE signal at the sampling times t, Ξ is the corresponding model accounting for R₂, R₂', and ΔB₀-induced signal modulations, $$$\Theta=\{Y_v,DBV,R_{2,nb}',\chi_{nb}\}$$$ is the unknown parameter set, ψ is the four-pool decomposition model for χ^9,10, Γ disentangles R₂' contributions from the non-blood and deoxygenated blood, and w and p are the regularization parameters that enforce prior knowledge in the qBOLD parameter estimation. Exact forms of Ξ, ψ, and Γ along with descriptions for sub-parameters therein are provided in Figure 2. In solving Eq. (1), VS-VSL-derived 3D CBV_v maps were employed to initialize and constrain the solution of DBV.

Experiments & data processing: Experiments were performed at 3T (Siemens Prisma) in six healthy subjects. Here, the AUSFIDE data collected in the study of Ref. 5 were used, while the VS-VSL data were newly acquired using the imaging parameters specified in Ref. 6. High-resolution T₁-weighted images were additionally acquired for brain segmentation. To evaluate the effectiveness of providing prior information of R₂' and CBV_v in solving the qBOLD problem, Eq. (1) was solved for an axial slice of a subject in the following three different ways: 1) p=0 and no CBV_v prior, 2) p=0 and CBV_v prior, and 3) p>0 and CBV_v prior, leading to Y_v and DBV maps, respectively. For the cases of 1) and 2), DBV was initialized with 2%, while for 3) DBV in each pixel was initialized to CBV_v with the solution bound: [0.5CBV_v, 2CBV_v]. χ was measured using the MEDI toolbox¹¹, and Y_v was initialized based on the χ values in the superior sagittal sinus. Hematocrit was measured individually for each subject, and was scaled with a factor of 0.85 for microvascular compartments¹². Derived 3D maps of Y_v, DBV, R'_2,nb, and χ_nb in a subject were presented in sagittal, coronal, and axial orientations. Group averages of Y_v and DBV across the six study subjects were calculated in gray and white matter regions.

Results

Figure 3 shows pairs of Y_v and DBV maps in the axial plane, derived in three different ways of using prior estimates of R₂' and CBV_v. Only the case employing both priors yields physiologically plausible Y_v (spatially nearly uniform) and DBV (gray and white matter contrast with the values in the former larger than those in the latter) maps. Figure 4 displays 3D maps of Y_v, DBV, R'_2,nb, and χ_nb in the three orthogonal planes. Group-averages of Y_v and DBV are Y_v=64.1±1.7% and 63.7±1.7%, and DBV=2.6±0.6% and 1.7±0.5% for gray and white matter regions, respectively.

Discussion and Conclusions

We introduced a new MRI oximetry method that enables 3D qBOLD parameter mapping across the entire brain. The results show the expected contrast in all four parameter maps obtained, i.e., a clear depiction of gray/white matter DBV difference, relatively homogeneous Y_v distribution. The data further highlight elevated R'_2,nband χ_nb in the iron-rich basal ganglia, relative to cortical regions, thus suggesting that the present method making use of preliminary estimates of R₂' and CBV_v achieves robust separation of several sources that contribute to voxel signal modulations in the brain. The proposed technique may find a range of clinical applications where oxygen extraction is elevated (such as in regional brain ischemia). Finally, in combination with cerebral blood flow imaging, the method should allow 3D voxelwise mapping of the cerebral metabolic rate of oxygen consumption across the entire brain.

Acknowledgements

This work was supported by the National Institutes of Health grant R21EB022687 and the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number UL1TR001878.

References

1. An HY, Lin WL. Quantitative measurements of cerebral blood oxygen saturation using magnetic resonance imaging. J Cereb Blood Flow Metab 2000; 20:1225-1236.

2. He X, Yablonskiy DA. Quantitative BOLD: mapping of human cerebral deoxygenated blood volume and oxygen extraction fraction: default state. Magn Reson Med. 2007;57:115-126.

3. Sedlacik J, Reichenbach JR. Validation of quantitative estimation of tissue oxygen extraction fraction and deoxygenated blood volume fraction in phantom and in vivo experiments by using MRI. Magn Reson Med. 2010;63:910-921.

4. Lee H, Englund EK, Wehrli FW. 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.

5. Lee H, Wehrli FW. Alternating unbalanced SSFP for 3D R2’ mapping of the human brain. Magn Reson Med. 2020. DOI: https://doi.org/10/1002/mrm.28637.

6. Lee H, Wehrli FW. Venous cerebral blood volume mapping in the whole brain using venous-spin-labeled 3D turbo spin echo. Magn Reson Med. 2020;84:1991-2003.

7. Yablonskiy DA, Sukstanskii AL, Luo J, Wang X. Voxel spread function method for correction of magnetic field inhomogeneity effects in quantitative gradient-echo-based MRI. Magnetic Resonance in Medicine 2013;70:1283-1292.

8. Han D, Nam Y, Gho SM, Kim DH. Volumetric R2* mapping using z-shim multi-echo gradient echo imaging. Magnetic Resonance in Medicine 2015;73:1164-1170.

9. Zhang J, Liu T, Gupta A, Spincemaille P, Nguyen TD, Wang Y. Quantitative mapping of cerebral metabolic rate of oxygen (CMRO2) using quantitative susceptibility mapping (QSM). Magn Reson Med. 2015;74(4):945-952.

10. Cho J, Kee Y, Spincemaille P, Nguyen TD, Zhang J, Gupta A, et al. Cerebral metabolic rate of oxygen (CMRO2) mapping by combining quantitative susceptibility mapping (QSM) and quantitative blood oxygenation level‐dependent imaging (qBOLD). Magn Reson Med. 2018;80(4):1595-604.

11. Liu J, Liu T, de Rochefort L, Ledoux J, Khalidov I, Chen W, Tsiouris AJ, Wisnieff C, Spincemaille P, Prince MR. Morphology enabled dipole inversion for quantitative susceptibility mapping using structural consistency between the magnitude image and the susceptibility map. Neuroimage 2012;59(3):2560-2568.

12. Eichling JO, Raichle ME, Grubb RL, Larson KB, Ter-Pogossian MM. In vivo determination of cerebral blood volume with radioactive oxygen-15 in the monkey. Circ Res. 1975;37:707-714.

Figures

Figure 1. Schematic of the proposed, preliminary estimates-based qBOLD parameter mapping procedure.

Figure 2. Descriptions for parameters pertinent to Eq. (1). For details of Ψ-related quantities, see Cho et al. (Ref. 10).

Figure 3. Pairs of Y_v and DBV maps obtained by solving Eq. (1) with three different usages of CBV_v and R₂′ prior information. Left: without CBVv and without R2’, middle: with CBVv and without R₂′, and right: with both CBVv and R₂′. Note that in the absence of either CBVv or R₂′ priors, both DBV and Y_v maps exhibit artifactual values in many voxels (left, middle), whereas in the presence of both CBV_v and R2′ priors, both the qBOLD problem becomes largely stabilized, leading to physiologically plausible values of Y_v and DBV (right).

Figure 4. Whole-brain 3D maps of Y_v, DBV, R′_2,nb, and χ_nb in the three orthogonal planes, obtained using the proposed qBOLD method. Note the expected contrast and physiologically plausible range in all four parameters, i.e., a clear depiction of gray/white matter differentiation, relatively homogeneous Y_v, and elevated R′_2,nb and χ_nb in the deep brain structures relative to cortical regions.

Proc. Intl. Soc. Mag. Reson. Med. 29 (2021)

0100