Knee ACL Tear, Meniscal Tear and Abnormality Detection Using Ensemble of CNN Techniques

Anterior cruciate ligament (ACL) and meniscal tears are the common injuries affecting sportspersons. These injuries affect the knee joints. The challenge of detection of these has been traditionally treated using image processing algorithms. Detection of ACL and meniscal tears has been approached using an ensemble-based CNN technique AMNet. The AMNet model consists of a pretrained ImageNet model ResNet50, XGBoost and voting classifier. A brief introduction of ACL, meniscal tears and the proposed approach is presented. The results of ACL tear, meniscal tear and abnormality detection using the pretrained CNN models, AMNet and other methods proposed have been compared for the performance metrics of accuracy, recall, precision and F1 score. The results strengthen the fact that AMNet outperforms the other models by detecting abnormalities with a mean accuracy of 89%, F1 score of 0.86, precision of 0.85 and recall of about 0.86.