Identification of luminal and medial adventitial borders in intravascular ultrasound images using level sets

Abstract

Extraction of the media and plaque boundaries from the intravascular Ultrasound (IVUS) images is gaining popularity as a biomedical application. This paper presents a novel system for the fully automatic extraction of the boundaries of the media and the plaque visible in the IVUS images. The system utilizes an enhanced level set technique to derive the evolution of two coupled contours as the zero level sets of a single higher dimensional surface. Moreover, the system utilizes the surface features to impose the expected media thickness. By using the single surface as a communication path between the contours, the system carries all the advantages of using two evolving surfaces and it becomes more efficient, less complex, easily extensible, and faster. Additionally, the capability of using different dynamic behaviors for the segmentation of the inner and outer walls makes our system even more flexible. The derived surface evolution equations capture the domain dependent information in an elegant and effective manner and address many practical issues, such as the missing wall sections or very weak boundary contrast. We have verified the accuracy and effectiveness of our system on synthetic and real data.