The use of simultaneous multiband radiofrequency (RF) pulses to accelerate volume coverage along the slice direction is becoming increasingly popular.1-3 To obtain higher resolution or higher signal noise ratio (SNR) maps, longer acquisition times and more brain volumes are necessary; in this context slice acceleration acquisition techniques are of great importance in diffusion tensor imaging (DTI), and for advanced encoding-intensive models such as diffusion spectrum imaging (DSI). However, there is limited agreement about the optimal protocol parameters for slice accelerated multiband acquisitions.4 In this work, we attempt to evaluate the impact on parameter calculations of parallel imaging in combination with multiband excitation for DTI applications.

Human brain experiments were performed on 8 subjects using a 3.0T MAGNETOM Trio (Siemens AG, Erlangen, Germany) with 32-channel head coil. Multiband RF pulses were generated for simultaneous multi-slice excitation and echo refocusing. Imaging parameters were as follows: TR / TE = 4000ms / 70ms, FOV=220 mm, matrix size = 110 × 110, voxel size = 2mm3, 52 slices, b-value = 1000 with 30 directions and 1 baseline. The slice-accelerated factor is set as 2, with two different in-plane parallel imaging (GRAPPA) factors: 1) iPAT = 1, NEX = 2, total time =2’44 × 2; 2) iPAT = 2, NEX = 4, total time = 3’00 × 4. All DWIs were processed in FSL including eddy current correction, motion correction and DTI indexe fitting, fractional anisotropy (FA) and mean diffusivity (MD) maps were calculated. Firstly, FA and MD from protocol IPAT 1 NEX 2 were compared with those from protocol IPAT 2 NEX 2. Voxel-based analysis and 2-sample T-test was applied here for quantitative comparison. Then the paired T-tests were conducted between 2 repeated acquisitions for iPAT 1 and iPAT2 protocols respectively.In figure 1, iPAT1 data showed higher SNR in DW and FA images (b = 1000) than iPAT2, especially for DW images in the central brain regions such as the basal ganglia, while its geometric distortion was more severe than iPAT2 (shown in B0 image). The voxel-based T-tests show the differences of FA and MD between iPAT1 and iPAT2 in Figure 2, with p < 0.001 and cluster size > 30. The iPAT1 data showed significantly higher FA and lower MD than iPAT2 data in basal ganglia region. The reproducibility evaluation showed that the FA and MD maps of both iPAT1 and iPAT 2 data show no significant differences (p< 0.001) between the two independent acquisitions.This experiment shows that the DTI indices based on slice accelerated multiband sequence are highly reproducible in voxel-based analysis for different parallel imaging factors, with no significant differences (p < 0.001). In addition, the parallel imaging factor may have an influence on SNR and distortion of the diffusion images. In detail, the IPAT1 data showed higher SNR especially at center brain region than IPAT2, however, it is more sensitive to susceptibility and eddy current distortion than IPAT 2. We expect the differences may be more significant with high b-value applications, such as DSI or other multiple b-value models. Because SNR may become an important factor in this applications, the choice of different parallel imaging factor should be considered depending on the application focus. The slice accelerated multiband sequence can significantly shorten the acquisition time and increase spatial resolution, which shows great impacts on neural studies. In this study, we evaluate the influence of parallel imaging based on the multiband sequence and hope the finding will guide our future applications of the sequence.