Automated Detection of Intracranial Aneurysms

Alexandra Lauric^a, Eric Miller^a,b, Sarah Frisken^a, Adel M. Malek^c,d

^a Tufts University, Department of Computer Science

^b Tufts University, Department of Electrical and Computer Engineering

^c Tufts University, School of Medicine

^d Tufts Medical Center, Department of Neurosurgery

The detection of brain aneurysms plays a key role in reducing the incidence of intracranial subarachnoid hemorrhage (SAH) which carries a high rate of morbidity and mortality. An intracranial aneurysm is a localized pathological dilatation of a blood vessel. It is reported that up to 2% of the general population harbors aneurysms. Most of these aneurysms are asymptomatic and remain undetected with only a small proportion proceeding to rupture and consequent SAH, with an annual incidence of approximately 1%. The majority of non-traumatic SAH cases is caused by ruptured intracranial aneurysms and accurate detection can decrease a significant proportion of misdiagnosed cases.

Although aneurysm detection is currently performed visually by experienced diagnosticians, there is an increasing interest in computed-aided diagnostic (CAD) systems for automated detection of intracranial aneurysms with the hope of improving the diagnostic accuracy and limiting missed detection.

A scheme for automated detection of intracranial aneurysms is proposed in this study. Applied to the segmented cerebral vasculature, the method detects aneurysms as suspect regions on the vascular tree, and is designed to assist diagnosticians with their interpretations and thus reduce missed detections. In the current approach, the vessels are segmented and their centerline is computed. Small regions along the vessels are inspected and a new surface descriptor is introduced to quantify how closely any given region approximates a tubular structure. Aneurysms are detected as non tubular regions of the vascular tree. The robustness of the method was investigated analytically and validated experimentally. The method was tested on 3D-rotational angiography (3D-RA) and computed tomography angiography (CTA). In our experiments, 100% sensitivity was achieved with average false positives rates as low as 0.66 per study.