Figure 4.

Qualitative comparison of the static and reference-guided SRR outcome of one subject for various input data scenarios in the sagittal view (additional axial and coronal view comparisons are shown in Supporting Information Figure S2). It illustrates the impact of the number of input stacks and how multiple orientations can improve PVE recovery. In particular, SRR (a+c+s+3obl) shows visually higher anatomical accuracy than SRR (2a+2c+2s) despite the same number of six input stacks used for the SRR. The red arrows (A) underline that the SRR based on only two stacks (a+c) as currently available for clinical MRCP study protocols produces a very poor SRR quality which is especially noticeable in the sagittal view. The magenta arrows (B) illustrate that for three input stacks (a+c+s) the corpus callosum can only be reconstructed with limited geometrical integrity. Motion correction helps to recover it more clearly by adding three additional stacks (2a+2c+2s) as indicated by arrows (C). The green arrows (D) show the improved visual clarity at the medulla due to better PVE correction using oblique data. Additional oversampling for high input stack numbers leads to higher PSNR. This may also result in clear tissue boundaries even in case of insufficient motion correction for the static SRR as indicated by the cyan arrow (E)