Blood volume fMRI with 3D-EPI-VASO: any benefits over SMS-VASO?
Laurentius Huber1, Dimo Ivanov2, Sean Marrett1, Puja Panwar1, Kamil Uludag2, Peter A Bandettini1, and Benedikt A Poser2

1Section of Functional Imaging Methods, National Institute of Mental Health, Bethesda, MD, United States, 2MBIC, Maastricht University, Maastricht, Netherlands

Synopsis

Cerebral blood volume (CBV) fMRI has the potential to overcome known limitations of BOLD fMRI with respect to spatial specificity and quantifiability of mapping brain activity. To overcome the coverage limitations of conventional CBV mapping with VASO, a novel VASO method with 3D-EPI readout was developed. This new approach is compared to BOLD fMRI and VASO with simultaneous multi-slice EPI readout. We provide evidence for a high sensitivity and improved specificity of 3D-EPI VASO compared to conventional BOLD fMRI. We conclude that because of its superior resolution in slice-direction, 3D-EPI VASO may play an important role in high-resolution fMRI.

Purpose

Quantitative cerebral blood volume (CBV) fMRI has the potential to overcome several limitations of BOLD fMRI. In particular, it helps to understand the underlying neuro-vascular coupling in positive and negative BOLD responses and offers superior localization specificity. Non-invasive CBV-fMRI with VASO, however, is inherently a single-slice approach and limited with respect to its field-of-view (FOV) [1]. This limitation can be overcome with advanced readout strategies such as simultaneous multi-slice (SMS) EPI [2] or 3D-EPI [3]. SMS-VASO has recently been used to obtain promising high-resolution CBV maps [4], however, with limited slice resolution. The purpose of this study is to develop and evaluate a new fMRI sampling strategy that combines VASO contrast with a 3D-EPI readout module. The performance of 3D-EPI-VASO is investigated with respect to its sensitivity and specificity to neural activity changes and compared with those of SMS-VASO and also standard BOLD fMRI.

Methods

Experiments were performed on a 7T Siemens scanner with a 32-channel NOVA head-coil. For simultaneous BOLD and CBV weighting, the SS-SI-VASO inversion preparation scheme was used [5]. We tested the new method in four pilot experiments of two volunteers. Sequence parameters of 3D-EPI and SMS were separately optimized to have the same FOV and were kept as similar as possible. Both approaches had the same TI1/TI2/TR=1.1/2.6/3.0s, bandwidth/Px=1860Hz, #slices=28, in-plane GRAPPA=2, in-plane partial-Fourier=7/8 and in-plane resolution=1.5×1.5mm2. To investigate the effective resolution in slice direction, the slice thickness was varied between 1.2-3mm in both SMS and 3D-EPI. SMS-specific parameters were: SMS-factor=4, blipped-CAIPI FOV-shift=1/3, TE=18ms, flip-angle=90°, excitation-pulse bandwidth-time-product=5.2. 3D-EPI-specific parameters were: through-plane GRAPPA=2, slice-partial-Fourier=7/8, TE=22 ms, flip-angle≈15°-25°, #segments=14, slab-selective excitation bandwidth-time-product=25. Please note the different acceleration for 3D-EPI and SMS-EPI, which is a result of advanced partial-Fourier imaging possible in 3D-EPI and the longer possible readout train within the VASO-specific timing constraints (TI/TR) with the 3D-k-space out-center-out (aka ‘linear’) trajectory. SMS-EPI unaliasing is done with split slice GRAPPA (MGH blipCAIPI C2P [2,6] (http://www.nmr.mgh.harvard.edu/software/c2p/sms). To investigate the specificity and sensitivity to neural activity changes, a 6-min finger tapping experiments (blocks of 30 s stimulation and 30 s rest) were conducted. Statistical activation and smoothness was estimated with FSL [7].

Results

Figure 1 shows the functional activation maps of CBV and BOLD acquired with 3D-EPI and SMS-EPI from one representative participant. The activation in CBV is on both sides of the central sulcus but not in between, while BOLD activation covers also CSF/vein in the sulcus (orange arrow in Fig. 1). 3D-EPI-VASO provides sufficient contrast-to-noise ratio to detect deactivation in the ipsilateral sensory cortex (blue arrows in Fig. 1) and the small excitatory activity in ipsilateral M1 (red arrows in Fig. 1), similar to BOLD fMRI. Representative maps of tSNR can be seen in Fig. 2. VASO tSNR in 3D-EPI is homogeneous across slices, while the different slice-dependent inversion times in SMS-VASO result in a heterogeneous tSNR (blue arrows in Fig. 2). Table 1 in Fig. 3 shows that 3D-EPI-VASO has approximately the same average tSNR compared to SMS-EPI, while yielding a higher resolution in the slice direction.

Discussion

The superior specificity of CBV-fMRI compared to conventional BOLD-fMRI is consistent with previous VASO studies [4,5]. The lower tSNR of 3D-EPI-VASO in the central imaging region is attributed to the lack of CAIPIRINHA-sampling in the 3D-EPI experiment compared to the blipped-CAIPI SMS-EPI (green arrows in Fig. 2). This results in higher g-factors, which can be addressed by 3D-CAIPI-EPI as previsouly shown [8,9]. The fact that VASO tSNR is comparable for 3D-EPI and SMS-EPI, but not in the BOLD signal, might arise from the different relative physiological and thermal noise contributions in the two contrasts, and the different spin history in BOLD and VASO imaging, when different flip angles are used in 3D-EPI and SMS-EPI. The higher slice resolution in 3D-EPI is consistent with slice profile imperfections in 2D-imaging. The proposed 3D-EPI VASO is currently being evaluated and optimized on a larger number of subjects, in order to more fully characterize its performance and inter-subject stability.

Conclusion

We combined VASO with 3D-EPI readout to increase the FOV comparable to conventional BOLD fMRI. 3D-EPI-VASO has superior localization specificity without the sensitivity to large draining veins within the central sulcus. Concurrently, it has enough sensitivity to detect small inhibitory areas in ipsilateral S1. 3D-EPI-VASO offers a higher resolution in the slice direction compared to SMS-VASO with comparable tSNR. Therefore, it may play an important role in high-resolution (layer-dependent) fMRI [5].

Acknowledgements

We thank Joelle Sarlls for her structure-phantom during sequence testing. We thank Steve Cauley for sharing the interface to his online SMS reconstruction on the scanner, which was used for the recon of the SMS-EPI VASO.

References

[1] Lu et al., MRM, 2003; 50:263-274; [2] Setsompop et al., MRM, 2012; 67:1210-1224; [3] Poser et al., NeuroImage, 2010; 51:261-266; [4] Huber et al., NeuroImage, 2015, doi:10.1016/j.neuroimage.2015.10.082; [5] Huber et al., NeuroImage, 2015,107; 23-33; [6] Cauley et al., MRM, 2013, 72:93-102; [7] Nichols, FMRIB TR08TNI, 2008; [8] Poser et al., ESMRMB, 2013, #287; [9] Narsude et al., MRM, 2015, doi:10.1002/mrm.25835.

Figures

Activation maps of VASO and BOLD with the 3D-EPI and SMS-EPI readouts. VASO has a superior specificity compared to BOLD. The different T1-contrast in the background of 3D-EPI compared to SMS-EPI images comes from the different spin history when different flip angles are used in 3D-EPI (20°) compared to SMS-EPI (90°). Activation maps are cluster-filtered with afni-3Dclust to exclude false-positive voxels despite low thresholds and limited tSNR.

tSNR maps of VASO and BOLD with 3D-EPI and SMS-EPI readouts for two different slice thicknesses. For BOLD, SMS-EPI seems to be superior compared to 3D-EPI. For VASO, the average tSNR is very similar, while SMS-EPI has a higher heterogeneity across slices (blue arrows). VASO has a lower tSNR compared to BOLD.

Table of GM tSNR values of BOLD and VASO images for all acquired readout strategies. The smoothness in SMS-EPI is inferior compared to 3D-EPI (last column).



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0944