INTRODUCTION

A computer-assisted diagnosis method for MRI is developed and used for different fields [1-3]. However, cruciate ligament rupture and anterior cruciate ligament (ACL) tears within the knee joint are hard to identify and few auto-assisted models and software are developed for diagnosis [4, 5]. In this paper, we introduce a fully-automated model of deep learning methods for identifying knee joint segment and distinguishing ACL tears. The scheme of our model consists of three CNNs including interested slides detection CNN, ligament detection CNN and ACL tear detection CNN. in the experiments, our model shows comparable diagnostic performance as human readers for detecting surgically confirmed ACL tears using sagittal IW-FSE images and T2-FSE images.