Rotator cuff tendon tears are one of the most common causes of shoulder pain and mobility limitations worldwide. Although ultrasound imaging is widely used for diagnosis due to its affordability and real-time capability, accurate interpretation heavily depends on the experience of radiologists. Variability in image quality, speckle noise, and unclear anatomical boundaries often make manual assessment challenging and time-consuming. This project presents an automated deep learning framework for detecting shoulder rotator cuff tendon tears from ultrasound images. The proposed system combines a Convolutional Neural Network (CNN) with an autoencoder-based contour segmentation approach to accurately identify key anatomical structures such as the humeral cortex and subacromial bursa. By focusing on meaningful structural boundaries instead of traditional pixel-wise segmentation, the model achieves improved precision and robustness. The segmented outputs are further utilized for classification using a deep CNN architecture to distinguish between intact and torn tendons. Experimental evaluation demonstrates strong segmentation accuracy and reliable classification performance, highlighting the potential of the proposed method as a supportive diagnostic tool. This system can assist clinicians in making faster, more consistent decisions while reducing dependency on manual interpretation. The approach represents a significant step toward explainable and efficient AI-driven medical image analysis in musculoskeletal ultrasound diagnostics. \\
Professor et al. (Thu,) studied this question.