부산대학교 도서관

Identification of an orthopedic Ankle Implant is highly important before deciding a revision surgery of joint replacement. The inability to identify an implant at the right time can lead to increased trauma to patients. The paper proposes a unique framework to identify the model and make of Ankle implants from radiography images of Anterior Posterior views using Deep learning. The work compares 3 different rigorously fine-tuned pretrained models to achieve the successful classification of implants. The use of the Inception V3 model provides the highest accuracy of 96.67% and an F1 score of 0.9662 in classification showing higher dominance in the accurate identification of Ankle implants.