A New Computer-Aided Solution for the Automatic Detection of Metal Stent Struts in Follow-up Evaluation in OCT Images

Stenting is commonly used in the treatment of coronary artery disease. To optimize the results of the procedure, it is crucial to evaluate the stent immediately after its implantation, and then to later monitor how it becomes covered by the neointima. One modality used for this is intravascular optical coherence tomography (IVOCT). While the identification and assessment of stent struts in IVOCT images is routinely done in clinical practice, manual analysis remains laborious and time-consuming. To address this, automated algorithms for stent segmentation have recently been developed. However, these are mainly used for stents without thick tissue coverage. This study proposes a new computer-aided method to automate the detection of both covered and uncovered metal stent struts in OCT images. In general, the algorithm involves segmenting potential stent strut shadows, analyzing the distribution of pixel intensities in detected areas, and then classifying objects. The algorithm has been validated on 606 cross-sections chosen randomly from 34 cases containing pullbacks: at baseline and at 3-36-month follow-up. The presented algorithm achieves sensitivity of 88% and precision of 90% including in-stent restenosis cases.