Preliminary Results for Automatic Detection of Arterio-Venous Malformations from Medical Images

Determining the blood flow patterns from medical images taken with a contrast agent is a problem situated at the congruency of multiple domains: medicine, image processing, mechanics and automatics. The outcome of such investigations can be used by clinicians in diagnosis and treatment of vascular diseases. In this paper we propose a method for automatic detection of a cerebral disease - the arterio-venous malformation (AVM) - using time series analysis performed on angiographies and their classification in three categories: arteries, veins and background. At the artery and veins merging, the AVM can be detected.