Virtual Brain Transplantation
Virtual brain transplantation
Because of mechanical deformation consequent to large cortical strokes, the anatomical parcellation on T1w images using semiautomated methods is very difficult to achieve. Hence, a “virtual brain transplant” process was performed in accordance with a previous approach (Solodkin et al., 2010). This method replaces the cortical lesion with the homologous image from the contralesional hemisphere from the same subject. With this, brain parcellation is possible using semiautomatized software. The process consisted of the following steps:
(1) Lesion segmentation by hand. (2) Using the AFNI 3dcalc function (Cox, 1996), the homologous region in the nonlesioned hemisphere was dissected and transplanted into the stroke region, effectively filling in the missing portions of the brain. (3) Manual corrections were then done in the interface between the native and transplanted T1-w images by visually examining each voxel and making voxel intensities uniform using AFNI’s 3dLocalStat and 3dcalc commands. (4) The brain was then parcellated into 96 cortical and subcortical regions. The original parcellation based on a macaque template (Van Essen, 2004) was transformed to the human MNI template via PALS (Van Essen, 2005). To increase accuracy, the deformation process was carried out using landmarks (based on CARET) and functional activation patterns considered homologous between the two species (Van Essen and Dierker, 2007).
ENEURO.0158-15.2016.pdf