<p>Anterior Cruciate Ligament (ACL) injuries are among the most common musculoskeletal conditions in athletes and physically active individuals, often leading to pain, instability, and long-term complications if not diagnosed early. Magnetic Resonance Imaging (MRI) is the gold standard for ACL evaluation, but manual interpretation requires expert knowledge and is prone to variability across radiologists. This study proposes an automated approach for ACL irregularity detection in MRI scans using the YOLOv8 deep learning model. A total of 284 annotated knee MRI images were collected and stratified into three categories: Normal, Partial Tear, and Complete Tear. The dataset was preprocessed through gamma correction, region-of-interest masking, and data augmentation, and the YOLOv8 model was fine-tuned via transfer learning to improve feature extraction and classification performance. Experimental evaluation on the test set demonstrated a mean Average Precision (mAP) of 0.91, with class-wise AP values of 0.94 for Normal, 0.91 for Partial Tear, and 0.89 for Complete Tear. Precision, recall, and F1-scores across all categories consistently exceeded 0.84, while the confusion matrix confirmed accurate classification with minimal misclassification between Partial and Complete Tears. These results highlight that YOLOv8 is capable of accurately localizing and classifying ACL abnormalities, offering radiologists a fast, reliable, and reproducible diagnostic aid. The findings suggest strong potential for integrating the proposed model into clinical workflows to improve diagnostic efficiency, reduce human error, and support timely treatment planning for ACL-related injuries.</p>