Multimodal brain images registration with severe pathological information missing

Clinicians regularly use several modalities to visualize different aspects of an explored organ. Indeed, recent medical imaging progresses allow complementary information often provided by anatomical and functional modalities to be combined. So, a preliminary image registration is required. In this paper, we present a registration method of multimodal brain MRI and SPECT images of a single patient. The constraint is that no external landmark are available. Therefore, we focus on seeking anatomical invariant brain properties. Hence, we explain our method based on fuzzy image segmentation and a principal axis computing on brain orientation considered as an invariant feature. We deal also with pathological cases presenting missing information in this context. An extension of our method based on a possibilistic shell fitting technique is then developed for brain missing information cases.