This study aims to utilize a novel convolutional neural network (CNN), MobileTurkerNeXt, to evaluate magnetic resonance imaging (MRI) for the diagnosis of SLAP and Bankart lesions of the shoulder. Images were taken from 343 patients, including those with Bankart lesions, SLAP lesions, and normal conditions. A majority were then augmented for training purposes by rotating images and adding noise, with the rest reserved for testing purposes. The 3 categories examined were Bankart vs normal, SLAP vs normal, and Bankart vs SLAP vs normal. Finally, the data was sorted into false positives, false negatives, and true positives/negatives. The CNN model was able to achieve an accuracy of over 96% in all 3 categories, matching true positives and negatives correctly each time. Comparison of other CNNs was also performed, with MobileTurkerNeXt significantly outperforming most other models in speed and accuracy. Researchers also integrated Grad-CAM-based interpretability, which is able to produce visual explanations that can support clinical assessment. Overall, MobileTurkerNeXt was demonstrated to be a highly effective CNN that has the potential to support clinicians within radiology, orthopedics, and other fields with diagnostic necessity.