Analysis of Atherosclerotic Plaques Using OCT Images Based on Deep Learning: A Comprehensive Review

Coronary atherosclerotic plaques are a leading cause of morbidity and mortality in coronary artery disease. Currently, optical coherence tomography (OCT) is employed by physicians to visualize intravascular plaques, a process that is labor-intensive and highly dependent on clinical judgment. Deep learning-based computer-aided techniques can effectively address these challenges. However, to date, no review has provided a comprehensive survey of these methods to direct future developments in the field. This review adopts two novel perspectives, namely the research task and the research question, which together address this gap in the literature by offering an in-depth discussion of 37 journal publications. Regarding tasks, this manuscript categorizes plaque research into three areas, which include detection, segmentation, and classification. In addition to analyzing and summarizing the benefits, drawbacks, and assessment metrics of deep learning models for each task, this section provides an overview of the issues in the field. The causes and solutions for the four challenges of noise, data shortage, data imbalance, and complex network structure are discussed in another topic. By providing a clear summary of the limitations in current research and suggesting future directions, this review aims to offer valuable insights to researchers working on coronary plaque analysis.