Medical image segmentation using information extracted from deformation

Deformation of normal structures in medical images has usually been considered as undesired and even a challenging issue to be tackled in medical image segmentation and registration tasks. With the objective of improving brain tumor segmentation accuracy in human brain magnetic resonance (MR) images, this paper proposes an approach to extract useful information from the correlation between lateral ventricular deformation and tumor. In some cases, comparative experiments show the improved tumor segmentation accuracy when the extracted information is added as an additional feature.