Automated Shoulder Implant Manufacturer Detection using Encoder Decoder based Classifier from X-ray Images.

Area of Science:

• Orthopedic surgery
• Medical imaging
• Artificial intelligence in healthcare

Background:

• Total shoulder arthroplasty (TSA) replaces shoulder joints with prostheses.
• Implant identification is crucial for revision surgeries but often hindered by poor record-keeping.
• Wear and tear necessitate potential future implant replacement.

Purpose of the Study:

• To develop an AI system for accurately classifying total shoulder arthroplasty implant manufacturers and models.
• To address the challenge of identifying implant types from X-ray images.
• To improve the efficiency of treatment planning for TSA patients.

Main Methods:

• An encoder-decoder based classifier was developed.
• Supervised contrastive loss function was utilized.
• The system was trained and evaluated using X-ray images of TSA implants.

Main Results:

• The AI system achieved 92% accuracy in identifying implant manufacturers.
• The method effectively overcame the class imbalance problem inherent in implant datasets.
• The proposed approach demonstrated high efficacy in classifying prosthesis types from radiographic images.

Conclusions:

• AI-powered classification of TSA implants from X-rays is feasible and accurate.
• This technology can significantly expedite the identification of implant models, aiding clinical decision-making.
• The developed AI system offers a promising solution for managing TSA implant information.