Journal of Networking Technology Lixin Wei 16 1 2025 https://doi.org/10.6025/jnt/2025/16/1/25-31 https://www.dline.info/jnt/fulltext/v16n1/jntv16n1_4.pdf By applying deep learning, we can better understand knee joint injuries that may occur in ball sports. This method can help us prevent these injuries more effectively and provide better protection for our athletes. This study developed a deep learning technique based on multiple modalities to identify common, major, and meniscus tear injuries, achieving rapid and accurate diagnosis. Additionally, it effectively categorizes the affected organs into different lesions based on their shape, size, and function. Experimental results demonstrated that the model accurately reflects the changes in the ROC curve, with an AUC change rate of 97.89% in normal anterior cruciate ligament tear or meniscus tear situations, significantly outperforming other methods and providing high accuracy.