Virtual brain transplantation (VBT): a method for accurate image registration and parcellation in large cortical...

The objective of the current study was to develop a semi-automated method to register and parcellate lesioned brains in a surface space with anatomical accuracy, facilitating group-level fMRI analyses in patients with large cortical strokes.

Thirteen chronic patients with a single large left hemisphere stroke were included in the study. Our "virtual brain transplantation" (VBT) approach is based on pre-processing high resolution anatomical T1-weighted brain images by "filling in" the lesion with "transplanted virtual tissue" from the non-stroke hemisphere, providing "normal" anatomical landmarks for standard alignment and inflation algorithms developed for healthy individuals. Biological validation of the approach was performed by quantifying in Freesurfer space the areas of 12 hand-drawn sulci found inside and outside the stroke following "transplantation".

Our results show no difference in the Freesurfer parcellation of 12 different regions when comparing a lesioned hemisphere with the non-lesioned hemisphere, attesting for the validity of the anatomical classification in the stroke hemisphere. As consequence of the anatomical precision, this method permits a more detailed and quantifiable anatomical description of the regions affected directly by the stroke.

This method permits accurate surface reconstruction of the injured hemisphere after stroke by making it possible to extract the cortical surface from these images and to utilize this in software modules (FreeSurfer) specialized for aligning cortical surfaces using high-dimensionality warping algorithms. In addition, it permits quantifying, within bounds, the extent to which the lesion in question is associated with damage to particular regions of the cortical surface, information that is of explanatory value in models that attempt to explain brain-behavior relations using lesion analysis.