Figure 1. Method schematics.

(A) 3D tumor ROI was identified from the anatomical MRI image. (B) The ROI was mirrored across the sagittal mid-plane to obtain the homologous region in the healthy hemisphere. A segmentation was performed and the intersection between the flipped tumor ROI and the white matter mask was computed. (C) The intersection mask was then transformed from the structural to the native diffusion space and used as the “seed” in the deterministic tractography analysis. Streamlines were generated from every point within the seed mask in the healthy hemisphere and two (D1) or only one (D2) target regions were identified; (E1) In the first case the WM bundles connecting the two target ROIs were investigated. The targets were sagittally mirrored and, after seeding, the tracts of interest in the lesioned hemisphere were generated in the same manner. (E2) In the second case the homologous region mask was dilated and the streamlines connecting the dilated ROI and the target were tracked to estimate the contralateral WM tracts involved in the tumor area. Subsequently, both the target and the dilated homologous ROI were sagittally flipped and the contralateral tracts were reconstructed in the pathological hemisphere. (E1)–(E2) The tracts of interest in the two hemispheres were finally defined in a probabilistic framework following the same tracking procedures and according to the number of targets identified. Single connection probability images were thus obtained to confirm deterministic tracking results.