Classification of Shoulder Impingement Syndrome using Transfer Learning model

Shoulder impingement syndrome refers to the inflammation and irritation of the rotator cuff. It refers to the insidious of sharp, anterolateral shoulder pain produced during elevation, eased on lowering the arm in the presence of a positive Neer and Hawkins Kennedy Test Tool utility, which typically occurs in patients over age 40. It is a common musculoskeletal medical condition affecting 7% to 26% of individuals before the Covid 19 and increases as individuals less physical activity during home quarantine. Over the last years, transfer learning methods have shown efficient results to assist radiologists and surgeons in classifying shoulder impingement syndrome. This abstract provides a brief overview of the most effective learning model used to classify shoulder impingement from the MURA-v1.1 dataset that is VRR16 and RESNET model. Although the shoulder impingement syndrome has been known for a long time, it remains an indisposed understood in the musculoskeletal medical entity. In contrast, hereby propose that regardless of the dataset size, we can use transfer learning model as baseline to gained knowledge previously to achieve highly accurate results. Figure model VGG16 results from predicted random four images Normal 79.99%, 96.89% while Abnormal 74.66% and 90.47%XRshoulder images. In comparison, the RESNET model has a more complex architecture compared with VGG16. Notwithstanding included efficient accuracy in classification, tuning hyperparameter and depending on epoch and batch size. Consequently, Figure model RESNET predicts four random images, likely the Normal 97.78%, 91.22% and Abnormal 81.36%, 78.62% XRShoulder images that achieve > 70% accuracy using deep learning technique such as transfer learning practical method with no clinical test use except for anonymous patient who volunteers to present medical image of shoulder impingement syndrome.