Self-supervised representation learning for detection of ACL tear injury in knee MR videos

• Novel CNN model is proposed for efficiently solving jigsaw puzzle as pretext task. The trained to learn explainable visual representational features. Proposed Divide-and-Teach strategy helps in resource-constrained training. It effective extract and transferable context-invariant First work on self-supervised learning ACL tear detection Knee MR. success of deep based models computer vision applications requires large scale human annotated data which are often expensive generate. Self-supervised learning, a subset unsupervised , handles this problem by meaningful features from unlabeled image or video data. In paper, we propose approach MR clips enforcing the anatomical task designed predict correct ordering jumbled patches that frames divided into. To best our knowledge, none supervised performing injury classification provide any explanation decisions made hence makes first its kind Experiments show enables spatial context invariant help reliable performance downstream tasks like Anterior Cruciate Ligament knee videos. efficiency novel Convolutional Neural Network paper reflected experimental results obtained achieves an accuracy 76.62% AUC score 0.848 Sagittal plane outperforming contrastive algorithms PIRL SimCLR using transformation. also achieved 0.740 KneeMRI dataset.