Shape-based assessment of intracranial aneurysm disease status

The risk assessment of intracranial aneurysms is an exceedingly difficult task. Clinicians associate aneurysm shape irregularity with disease instability. However, there is no consensus on which shape features reliably predict aneurysm instability. We have adopted a machine learning approach to identify shape features with predictive power for aneurysm instability: From imaging data 3D models of aneurysms are extracted that are used to train a classifier. A variety of representations of the 3D shape are calculated, these include the Zernike moment invariants (ZMI) and geometry indices such as aspect ratio, ellipticity and non-sphericity. The processing pipeline was applied to synthetic data and clinical datasets of 413 aneurysms registered in the AneurysmDataBase (SwissNeuroFoundation) and AneuriskWeb database. Classification based on ZMI alone allowed us to distinguish between sidewall and bifurcation aneurysms, but failed to forecast an aneurysm’s rupture status reliably. Simpler geometry indices performed similarly well in rupture status prediction. On synthetic data we showed that ZMI could encode shape irregularity. It remains to be investigated whether further stratification of the aneurysms in terms of location, size and clinical factors will increase the robustness of the applied classification methods