Abstract: A 3D parallel method is proposed for segmenting intravascular ultrasound (IVUS) image sequence to extract vessel wall borders based on snake model. Firstly, original images are preprocessed to reduce possible noises and suppress ring-down artifacts. Then, four IVUS longitudinal cuts are obtained where lumen and media-adventitia borders are detected. Once these borders are mapped onto each cross-sectional slice, initial contours in each IVUS frame are obtained. Finally, these initial contours evolve continuously through minimization of a snake energy function until stop at target borders. Consequently, segmentation of each IVUS frame is implemented simultaneously and the efficiency is greatly raised compared with 2D sequential segmentation approaches. The proposed method is experimentally evaluated in large datasets of IVUS images clinically derived from human coronary arteries. Results demonstrate that IVUS images can be automatically, quickly and reliably segmented with the method.

Key words: computer application, image segmentation, parallel processing, intravascular ultrasound (IVUS), snake model