GRAPPA reconstructed wave-CAIPI MP-RAGE at 7 Tesla

FIGURE S1 Comparison of 16-fold accelerated wave-CAIPI (ACS included) and 12-fold accelerated Cartesian CAIPIRINHA (ACS excluded). Accelerated data were simulated from the fully sampled datasets and the reconstruction was performed using 16 × 16, 20 × 20, 24 × 24, and 32 × 32 ACS lines for the computation of the GRAPPA weights. An axial view of the reconstruction and the difference to the fully sampled image (scaled 10×) are shown. The RMSE in the brain region and the acquisition time of a respectively accelerated scan are given. The RMSE averaged over all 5 subjects is depicted in brackets. The yellow line indicates the ROI used for the RMSE. The results show, that highly accelerated Cartesian CAIPIRINHA reconstructions are more prone to image artifacts, especially for smaller ACS sizes. 16-fold accelerated wave-CAIPI performs better than an approximately time-matched 12-fold accelerated Cartesian CAIPIRINHA protocol with 20 × 20 ACS lines and external reference scan. Additionally, wave-CAIPI allows a further scan time reduction by acquiring only 20 × 20 or 24 × 24 ACS lines while maintaining good image quality

FIGURE S2 Axial view of the automatic subcortical segmentation achieved within FreeSurfer's longitudinal processing stream. The results are shown for one of the subjects using Cartesian CAIPIRINHA and wave-CAIPI sampling with different acceleration factors

FIGURE S3 SENSE-type wave-CAIPI reconstruction of retrospectively 16-fold accelerated wave-CAIPI acquisition. Different ESPIRiT thresholds were used for the computation of different sizes of sensitivity masks (right). The RMSE with regard to the fully sampled data was calculated in the brain region for all three reconstructions. The default mask (threshold = 0.8) computed with the ESPIRiT algorithm crops out the mouth and nasal cavity region and thereby, the reconstruction slightly falls of in quality. The smallest RMSE was found for the threshold 0.4, however, the results are very similar