Unsupervised MR harmonization by learning disentangled representations using information bottleneck theory

Lianrui Zuo, Blake E. Dewey, Yihao Liu, Yufan He, Scott D. Newsome, Ellen M. Mowry, Susan M. Resnick, Jerry L. Prince, Aaron Carass

Abstract: Highlights•Unsupervised MR harmonization without traveling subjects.•Unified latent space for MR contrast synthesis.•A novel framework for disentangling contrast and anatomy in MR images.•Downstream segmentation consistency shows significant improvements after harmonization.