Harnessing the Potential of Deep Learning for Total Shoulder Implant Classification: A Comparative Study

Aakriti Mishra, A. Ramanathan, Vineet Batta, C. Malathy, Soumya Snigdha Kundu, M. Gayathri, D. Vathana, Srinath Kamineni

Abstract: Orthopedic implant identification is an important and necessary step prior to performing revision surgery of different joints. The inability to identify an implant can lead to significant surgical difficulties with consequent unfavorable outcomes. This paper proposes a novel framework to identify the make and model of seven (7) different total shoulder arthroplasty implants utilizing plain X-ray images and Artificial intelligence. The proposed work classified implants with an accuracy of 91.48% and with an AUC (Area under curve) of 0.9932 showing higher effectiveness in orthopedic implant identification. Further work is required to enhance and progress this work, with a goal of greater accuracy and fewer errors.