Computer-Aided Diagnosis of Cerebral Aneurysm Based on Fuzzy Expert System: MR Angiography Study

MRA (magnetic resonance angiography) is a commonly used method for diagnosing unruptured cerebral aneurysms. In this paper, we introduce a novel computer-aided diagnosis (CAD) system for the cerebral aneurysm using MRA images. The proposed CAD system automatically finds candidates of the cerebral aneurysm, and then evaluates a fuzzy degree belonging to aneurysm whose size is more than a user-specified size. To find the aneurysm candidates, the normal model, which has no aneurysm, is estimated by using the cerebral arteries segmented from MRA images. Then the candidates are characterized by four features; variance, hemi-sphericity, mean MR signal value, and directional gradient concentration. The system then estimates the fuzzy degrees by using a feature value map generated by reference datasets. In the experiments, the proposed system was applied to eight patients with cerebral aneurysms. The four patients were used to make reference datasets, and the four patients were used to test the ability of finding cerebral aneurysms. The experimental results showed that our system gave the highest fuzzy degree for the cerebral aneurysm among the candidates in all cases except one case with small aneurysm.