Automatic Lumen Segmentation in Intravascular Optical Coherence Tomography Using Morphological Features
Huishuo Zhao,Bin He,Zhenyang Ding,Kuiyuan Tao,Tianduo Lai,Hao Kuang,Rui Liu,Xiaoguo Zhang,Yicheng Zheng,Junyi Zheng,Tiegen Liu
DOI: https://doi.org/10.1109/access.2019.2925917
IF: 3.9
2019-01-01
IEEE Access
Abstract:Lumen segmentation in intravascular optical coherence tomography (IVOCT) images is a fundamental work for more advanced plaque analysis, stent recognition, fractional flow reserve (FFR) assessment, and so on. However, the catheter, guide-wire, inadequate blood clearance, and other factors will impact on the accuracy of lumen segmentation. We present a simple and effective method for automatic lumen segmentation method in IVOCT based on morphological features. We use image enhancement, median filtering, image binarization, and morphological closing operation to reduce speckle noise, minimize the effect of blood artifacts and fill in small holes inside vascular walls. We extract the orientation and area-size of connected regions as morphological features in images and remove the catheter and guide-wire completely by morphological corrosion operation, small area-size region removal, and orientation morphological feature comparison, and then the contour of the lumen can be discriminated. The evaluation metrics of this method, the Dice index, Hausdorff distance, Jaccard index, and accuracy of 99.32%, 0.06 mm, 99.4%, and 99.66%, respectively, are obtained from comparing with expert annotations on 268 IVOCT images. Compared with the other morphology-based lumen segmentation methods, the presented method can remove the catheter and guide-wire completely, even if the catheter and guide-wire cling to the lumen or the shape of the catheter is irregular. Since only morphological operations are used to complete all processes, the calculation burden is reduced greatly.