Rigid Lens - Locally Rigid Approximations of Deformable Registration for Change Assessment in Thorax-Abdomen CT Follow-Up Scans

摘要

A general problem of any deformable image registration method for change assessment is to find a good balance between computing a precise match and keeping locally differences. In this work we present the rigid lens concept dealing with this issue. The rigid lens is based on locally rigid approximation of locally precise deformations and can be used for interactive viewing and visualization of changes as well as for automatic change detection. We demonstrate the rigid lens in the context of oncological workup of thorax-abdomen CT follow-up scans and evaluate the concept for change assessment based on a study with 1492 manually annotated lesion in scans from more than 400 patients.