Machine learning for anterior cruciate ligament and meniscus analysis in knee MRI: A comprehensive review

Abstract

Anterior Cruciate Ligament (ACL) and meniscal injuries are prevalent in knee joint problems and are closely associated with osteoarthritis. With the rapid development of machine learning (ML) in knee magnetic resonance imaging (MRI) interpretation, a surge of research on ACL and meniscus analysis has emerged. However, there has been a noticeable absence of comprehensive reviews that can offer detailed classification, analysis, and comparison of the rich existing methods. To fill this gap, we provide an overview of ML methods applied in ACL and meniscus MRI analysis between 2013 and 2024. Sixty-seven papers covering tasks such as classification, segmentation, localization, and prediction are investigated and classified from the perspective of the ML method. For conventional ML methods, we summarized four kinds of handcrafted MRI features related to the ACL and meniscus, along with corresponding ML models. Meanwhile, we categorize the deep learning methods into 11 types according to the network structures for various tasks. Based on the categorization, we further compare the main methods and analyze the critical factors for existing successful models. Current trends and future directions in this field are also elaborately discussed. Serving as a technical reference, this survey hopes to inspire researchers interested in this field in method selection and design, which ultimately advances ML in clinical applications of ACL and meniscus.