Multiple Area B1 Shimming: An efficient, low SAR approach for T2-weighted fMRI acquired in the Visual and Motor Cortices of the Human Brain at Ultra-High Field

Introduction/Synopsis B1 heterogeneities are a major challenge at Ultra-High Field[1,2]. B1 shim techniques can mitigate those inhomogeneities but B1 Shim solutions aiming at uniform B1 over the whole brain generally result in poor RF efficiency because of large destructive interferences [3]. This translates into higher RF power, thus higher SAR levels. Less constraining tradeoffs can be obtained, sacrificing to some degree on B1 homogeneity, but some sequences are especially sensitive to flip angle variations. This is the case for a T2 weighted sequence that was previously developed [4] in order to obtain multi-slice T2w fMRI at 7T with low levels of SAR. The critical T2 preparation of this sequence cannot be efficient if large B1 variations occur within a preparation slab. Here we demonstrate that a multi-region B1 Shim approach allows for improving T2 contrast in different regions of the human brain (visual and motor cortices) with better homogeneity and lower RF power. Methods A slab wise magnetization Preparation for Functional Imaging with a T2 weight (SPIF-T2) [4] is used to provide the T2 weighting for the more accurate Spin Echo (SE) fMRI [4-6], while reducing SAR significantly (~3 fold for 10 slices when compared to a standard multi slice Spin Echo (SE) sequence). Ten slices were positioned to go through either the visualor the motor-cortex. This technique is used in conjunction with Parallel Imaging with X4 acceleration (1D) and a half-Fourier technique to allow for whole brain coverage while maintaining short acquisition times necessary to keep Gradient Echo (GE) contributions small. One normal subject participated in this study. Experiments were performed on a 7T system (Siemens, Magnex). The motor (finger tapping) and visual (flashing red checker board) paradigms consisted of 10 blocks of 30s stimulus and 30s rest with a total duration of about 10 minutes. Each 30s period consisted of 5 acquisitions. Each acquisition consisted of the same T2 prepared 120 mm slab going through the visualor the motor-cortex. The slab selective T2 magnetization preparation, consisted of a (90o |180o|-90o) RF sequence to flip back the magnetization along the z axis, followed by 10, interleaved GE EPI slices of 2mm thickness each. (FOV=19.2x19.2cm; matrix=128x128, single shot; α=90°); TE for the preparation slab was 55 ms; TE for the EPI readout with half-Fourier was 5.9 ms. TR in the multi slice EPI train was ~ 25 ms per slice leading to 250 ms for the 10 slice acquisition following each T2 preparation module; this includes a 12.2 ms fat suppression module for each slice). For comparison, a similar dataset but without the T2 preparation module was obtained. Identical readout was played in prepared and nonprepared acquisitions except for a smaller flip angle in the non prepared case to account for the reduced SNR due to the T2-weighing in the prepared case. Two B1 shim targets were defined based on two axial slices positioned in the center of the two slabs chosen for the subsequent fMRI series (one in the visual cortex, one in the motor cortex). Within each of these two axial reference slices an ROI was drawn defining the B1 shim target location. It was sufficient to utilize the center slice as transmit B1 varied only slowly along the Z direction, i.e. around those target slices. A 3D B1 Map of the whole brain was obtained with the AFI technique with a nominal flip angle of 70 degrees [8]. For each B1 shim target a series of 18 GE images were obtained with a small flip angle (16 images one channel transmitting at a time, one image all coils transmitting, one image without pulsing RF) to produce relative B1 maps [3]. Those relative maps merged with the 3D B1 map yielded 16 magnitude and phase B1 maps for each channel [9]). A B1 shim solution was calculated for each target using the optimization toolbox in matlab. 3D B1 maps were measured again with the two B1 shim settings to validate the predicted B1 alterations. Note that only B1 phase modulation was utilized in the non linear optimization algorithm. RF power rescaling was performed in a second, optional step.