Introduction

Venous cerebral blood volume (CBV_V) is one of important parameters for understanding the underlying mechanism of BOLD MRI signals¹. Yet, imaging-based methods enabling voxel-wise quantification of CBV_V are sparse², due primarily to a very small portion of venous blood in tissues and difficulty in targeting the venous compartment exclusively. Recently, the velocity-selective venous-spin-labeling (VS-VSL)-prepared 3D turbo spin echo (TSE) method³ has shown promise with that regard, achieving whole-brain CBV_V mapping based purely on endogenous contrast mechanism. Nevertheless, the method’s relatively long scan times (~3 minutes) limits its application to functional studies involving short-term stimuli. Thus, in this work we aimed to develop a highly accelerated technique for CBV_V mapping across the entire brain.

Methods

Sequence configuration: Figure 1 shows a timing diagram of the proposed pulse sequence, which consists of three consecutive modules: 1) magnetization preparations (Fig. 1a) leading to selective nulling of arterial blood and cerebrospinal fluid signals, 2) velocity-selective (VS) RF pulse train (Fig.1b) for exclusive labeling of venous blood spins, and 3) 3D gradient-and-spin-echo (GRASE) encoding (Fig. 1c). To accelerate data acquisitions relative to the parent method³, a particular focus in this work was given to the GRASE module. Specifically, variable flip angles (Fig. 2a), calculated based on a prescribed two-step signal evolution (Fig. 2b), were applied in the refocusing RF pulse train so as to achieve a long echo train length and thus short scan times. Phase-encoding views are sampled in a sparse elliptical ky-kz space, leading to further scan acceleration. Here, a segmented linear, center-out view-ordering scheme⁴ was employed (Figs. 2c, 2d) to minimize a loss of efficiency in venous blood spin labeling, while preventing image artifacts potentially resulting from inter-echo signal inconsistency.

CBV_V mapping: Control (VS gradients off) and tag (VS gradients on) data acquired with sparse and incoherent sampling in k-space were individually processed using the standard compressed sensing image reconstruction technique⁶. Based on the assumption that T2 values of brain tissues and venous blood are in the same range at 3 T field strength, CBV_V quantification can be simplified to the following equation³:

$$ \frac{S_{control} - S_{tag}}{S_{control}} = CBV_V $$

Experiments and data analysis: Experiments were performed at a 3T scanner (Siemens Skyra) in four healthy subjects using VS-VSL-GRASE with the following imaging parameters: field-of-view: 220 x 220 x 120 mm³, reconstruction matrix size: 72 x 72 x 60, sagittal orientation, echo train length = 36, GRASE factor = 3, RF spacing = 5.2 ms, and gradient-echo spacing = 1.2 ms. To ascertain the method’s maximally achievable acceleration factor without loss of quantification accuracy, data were acquired with three different sampling ratios: full elliptical ky-kz sampling, 50 %, and 30 %, leading to scan times of 185, 123, and 74 seconds, respectively. Additional data were collected in each subject using high-resolution MP-RAGE for brain segmentation. Whole-brain VS-VSL images along with derived CBV_V maps obtained with full elliptical k-space sampling were reformatted in sagittal, coronal, and axial orientations. Whole-brain CBV_V maps for the three cases with different sampling ratios were also presented. Furthermore, regional averages of CBV_V were computed in gray and white matter for all experiments performed.

Results

Figure 3 shows whole-brain 3D images in the three orthogonal planes in a study subject: VS-VSL control (Fig. 3a), tag (Fig. 3b), control/tag difference (Fig. 3c), and CBV_V (Fig. 3d). The difference image highlighting large veins suggests that the venous blood spins are properly labeled as intended, while CBV_V maps exhibit the expected contrast between gray and white matter regions. CBV_V maps in Figure 4 obtained with different data subsampling ratios are visually comparable. Table 1 lists CBV_V averaged across the entire gray and white matter voxels in the four study participants scanned at various sampling rates.

Discussion and Conclusion

We introduce a scan-time efficient CBV_V mapping strategy by means of GRASE encoding with sparse k-space sampling. Compared with the parent method based on TSE readout, the presented technique achieves scan accelerations up to approximately three-fold. While resulting CBV_V maps depict physiologically known contrast within plausible ranges (group averages: gray matter ~ 2% and white matter ~ 1%), the values are somewhat slightly lower than those in previous reports^{3, 5}. Thus, we are currently focused on probing the cause of this difference while optimizing the sequence and scan parameters to explore a possibility of further reduction of imaging time. Upon further evaluation of these issues, the present GRASE-based CBV_V mapping method is expected to find various applications in neuroimaging studies.

Acknowledgements

No acknowledgement found.