Graph based lumen segmentation in optical coherence tomography images

Intravascular optical coherence tomography (IVOCT) is a new invasive imaging system which produces high-resolution images of coronary arteries. Lumen segmentation plays an important role in subsequent analysis of IVOCT images. In this work, we develop a fully automatic lumen segmentation method on IVOCT images. A graph based method is applied to segment the vessel lumen and a match filter based method is employed to detect the guide-wire artifact which caused by guide-wire. A dataset of 500 IVOCT images with manually labeled lumen boundaries is used to evaluate the proposed approach. Overlap dice (OD) is computed to quantitatively evaluate the segmentation result. Results show that the proposed graph based segmentation method is accurate and efficient.