This event is for the Stanford community.
“Identification of ACL and Meniscal Pathology Using Machine Learning”
Machine learning methods for image segmentation and feature identification have rapidly advanced over the past several years. However, application of these methods to MR exams of the knee remains challenging due to the joint’s anatomic complexity as well as the requirement for large amounts of accurately-annotated training data to achieve high accuracy. In this work, we restrict our attention to specific, clinically significant structures – the menisci and anterior cruciate ligament. We generate an in-house training data set comprised of normal knees, knees with meniscal/ACL pathology, and knees with pathology outside these structures. The menisci and ACL are segmented using a semi-automated process which partially preserves the anatomy of the adjacent knee joints. Using this data, we investigate the performance of a machine learning-based method to identify and classify ACL and meniscal pathology.