Recently, Optical Coherence Tomography (OCT) has become one of the preferred clinical techniques for intracoronary
diagnostic imaging. Thanks to its high resolution imaging capability, the OCT technique allows to identify microscopic
features associated with various types of coronary plaque and to track of stent position, malaposition and neo-intimal
tissue growth after stent implantation. Accurate visualization of stent struts can help to examine the status of implanted stents potentially leading to proper treatment of the coronary artery disease. However, unfortunately, current stent identification involves time-consuming segmentation algorithms sometimes requiring labor-intensive manual analysis process. To resolve the problem, we propose a high-speed automatic segmentation algorithm of intravascular stent struts in OCT images. Unlike the other "automatic" stent segmentation algorithms, which are mainly based on time-consuming machine learning algorithms with manual addition and removal of stent struts for correction during the analysis process, our algorithm does not require any manual adjustments of stent struts. Our algorithm first analyzes 10 consecutive crosssectional OCT images to take boundary information into account to enhance the accuracy of guide-wire segmentation and lumen segmentation. Then, it performs stent segmentation by automatically eliminating guide-wire signals using the previous segmentation results. The implementation of our algorithm uses the Intel(R) IPP library on CPU and the CUDA technology on GPU, which achieves the average analysis time of 0.28 s/frame and the detection rate ranging from 84% to 88.6% for about 120 continuous images per patient. As such, the proposed algorithm is robust and fast enough to be integrated in clinical routine.