Skip to content
Chris Rorden edited this page Nov 4, 2017 · 1 revision

enantiomorphic normalization (20171104)

Normalization warps each individuals brain so they have the same shape. We do this by adjusting the shape of the persons brain to match a stereotypical template image. This automated process can be disrupted by the presence of a brain lesion (which is visible on the individuals scan but not present on the template). We use enantiomorphic normalization to solve this problem. This method leverages the symmetry of the human brain and the fact that most brain injuries only impact one side of the brain. By aligning the individual's scan with their mirror image scan we can make a 'healed' scan where injured tissue is replaced with tissue from the healthy hemisphere. However, not everyone has perfectly symmetrical brains. Previous solutions have used rigid body coregistration to align the hemispheres. However, this sometimes fails for individuals with very asymmetric brains. The image below shows an individual the most asymmetric individual observed from 181 people. Note the rigid body coregistration (green, 6 degree of freedom) does not match the brain's outline well. Therefore, our latest software includes a full affine (12 degree of freedom) transformation, which results in a much better fit (red). Note that non-linear transforms would not be appropriate as these would tend to shrink the volume of the lesion.

Clone this wiki locally