Automatic Brain Aneurysm Extraction in Angiography Videos using Circlet Transform and a Modified Level Set Model

Abstract

Background: These days, many attempts have been done to specify the size and location of aneurysms, leading to more successful surgical operation and less bleeding risk. In this paper, a novel method is proposed to extract brain aneurysms from two dimensional x-ray angiography videos, automatically. Methods: The most acute challenges in detecting brain aneurysm are the complexity of vessel structures and shape similarity between the aneurysm and vessel overlaps and vessel cross sections. Therefore, researchers regarded removing vessel structures as an initial and crucial step to detect aneurysm. Since the circularity feature is the most distinctive criteria for physicians to detect aneurysm, firstly, we proposed a robust method based on Fast Circlet Transform (FCT) to localize the aneurysm without needing to remove vessel structures. Then, to segment the detected aneurysm more accurately, a modified Level Set algorithm is proposed. Finally, our proposed method is quantitatively evaluated on two different datasets with different views, shapes, sizes, locations and contrast. Results & Conclusion: Experimental results show that the proposed system is reliable without dealing with vessel structure removal challenges, reluctant false positive candidates, hard parameter tuning and poor edge gradient.