Analysis of cerebral ischemia MRI images using non-rigid registration

Abstract

In diagnosis and treatment of brain diseases, doctors often need to observe the lesions at different time to evaluate the changing conditions of the diseases or judge the effect of treatment. In this paper, we employ two non-rigid registration methods to match MR images of cerebral ischemia in different periods, including a non-rigid registration algorithm based on free-form deformation (FFD) model and a DEMONS registration algorithm which is used to be compared with the FFD method. The results of our experiments suggest that non-rigid registration works well on realistic data of brain lesions with small deformation (k=0.5∼0.8) and the cross correlation coefficients (CC) increases from 0.406 before registration to 0.683 after FFD registration and 0.728 after DEMONS registration.