Automatic stent strut detection in intravascular ultrasound using feature extraction and classification technique

The Bioabsorbable Vascular Scaffold (BVS)is a temporary stent, which provides support to the vascular lumens and then gradually resorbs over time to improve the recovery of blood flow in the blocked vessel. The recognition of BVS after implant is a widely used procedure in the clinical management of coronary artery disease to estimate the effects of the treatment. However, manual recognition process is time consuming, tedious and prone to human errors. This paper proposes a new computer aided solution to automatically identify and mark the stent strut in intravascular ultrasound images. We use AdaBoost based ensemble learning approach to accommodate various classifiers from different methods to enhance the performance. We can use simple features to filter, and there is no over-fit phenomenon. Moreover, Support Vector Machines (SVM)algorithm performs high efficiency, and significantly improves the learning accuracy. During the training process, images are normalized and feature extraction is carried out to train a cascaded AdaBoost classifier and a SVM classifier. Then the recognition images are output with identified objects in the testing process by using features and classifiers obtained from training. The proposed approach significantly guarantees both the precision and computational efficiency, and can be widely applied in the clinic to facilitate stent recognition and visualization potentially, adding stent implantation.