Multi-echo Parallel Imaging Accelerated fMRI with Susceptibility-induced Off-resonance Compensation

A 1.5T GE Signa 12X Excite unit with an 8-channel phased array head receiver coil (Invivo Cooperation) was used for all measurements. TE’s were set to 23ms, 48.9ms, and 74.8ms. A flip angle of 70° and a TR of 1500ms were applied. The slice-select rephasing gradient was set to 30% of its original area. A second refocusing gradient in z-direction was applied following the first echo train to fully rephase the spins in regions without off-resonance effects. 15 5mm-slices were acquired. A breath-holding experiment was performed by the volunteer to stimulate functional activation in all areas of the brain. A baseline period of 16.5s was followed by a 3s-long breath-in period and a 16.5s long breath-holding period. The on-/off-cycle was repeated eight times, followed by a 18s long off-period, and resulting in a total scan time of 5:06 min. The correlation of the stimulus response with a sinusoid function [7] was calculated for the first echo image and the average of the second and third echo images separately, resulting in two separate correlation maps. The correlation coefficient of the first echo image was considered for all voxels in which the signal magnitude of the first echo image was higher than the corresponding magnitude of the second echo image. For all other voxels, the values derived from echo two and three were used to create the final activation map. Finally, a correlation threshold of r=0.3 was applied, above which a particular voxel was considered activated by the stimulus. Results Fig.2 shows slice images from one time point in the fMRI experiment. In the non-compensated echo images (Fig.2b/c), dropout regions are clearly visible, while the signal from these regions was recovered in the susceptibility-compensated first echo image (Fig.2a). Signal dropouts were almost eliminated in the combined root-sum-ofsquares images (Fig.2d). Fig.3 shows the resulting fMRI activation maps from a) the susceptibility-compensated method and b) a separate 3-echo fMRI experiment using a regular rephasing gradient (no z-shimming). The fMRI signal could be successfully restored in the aforementioned dropout regions (encircled in green in the noncompensated experiment). Discussion The results clearly show restored signal in typical dropout regions usually “invisible” to the scanner. In terms of the fMRI experiment, activation could be restored in these regions. In more uniform regions, the fMRI signal was not compromised by the lack of signal from the first echo image. Therefore, our susceptibility-compensated multi-echo fMRI acquisition technique can be used to acquire fMRI data with whole-brain coverage without significant reduction in sensitivity. The averaging of two echo images (echoes two and three) restores most of the sensitivity that was lost by parallel imaging. Furthermore, this technique does not prolong scan time compared to standard acquisition techniques, making it a valuable alternative for whole-brain fMRI measurements where dropout regions are a concern.