Using shape descriptors to categorize intracranial aneurysms

Working towards a system to match intracranial aneurysms of similar morphology, we examine how a variety of established shape descriptors relate to expert assessments of aneurysm morphology. For this purpose, we extract 3D models of aneurysms from 3D angiographies and calculate numerical values (geometry indices and moment invariants) that describe the size and shape of the aneurysm and the surrounding arteries. We then compare these shape descriptors with human assessments of qualitative morphological characteristics like irregularity or the presence/absence of blebs. Here, we present preliminary results based on 137 aneurysm models.