A Novel Segmentation Approach for Intravascular Ultrasound Images
Intravascular ultrasound (IVUS) is an important imaging technique to study the architecture of vascular wall for diagnosis and assessment of vascular diseases. Segmentation of lumen and media-adventitia boundaries from IVUS images is the foundation for quantitative assessment of the vascular walls. In this paper, a novel and fully automated segmentation approach is proposed for IVUS images. The proposed approach utilizes different segmentation strategies for lumen and media-adventitia boundaries respectively according to their different region and boundary features on IVUS images. For lumen, the segmentation is carried out by combining image gradient with a fuzzy connectedness model. For media-adventitia boundary segmentation, minimal path based on fast marching model is used. The effectiveness of the proposed IVUS image segmentation approach was validated by segmenting 180 IVUS images from nine patients, in which the results were compared to the corresponding gold standard. The approach achieved high mean overlap area ratios of (86.49 ± 16.93%)/(92.73 ± 5.47%) and small average boundary distances of (0.09 ± 0.10) mm/(0.07 ± 0.06) mm for lumen and media-adventitia, respectively. The obtained results have shown the potential of the proposed approach to effectively segment lumen and media-adventitia boundaries from IVUS images.