Development of a machine learning algorithm to identify total and reverse shoulder arthroplasty implants from X-ray...

Area of Science:

• Orthopedic surgery
• Artificial intelligence
• Medical imaging

Background:

• Total shoulder arthroplasty (TSA) demand is rising, with reverse TSA (rTSA) incidence increasing significantly.
• Timely identification of shoulder implants is crucial to minimize operative time, costs, and complications.
• Machine learning (ML) applications in shoulder surgery remain limited despite their potential in medical image analysis.

Purpose of the Study:

• To develop and evaluate a machine learning model for identifying shoulder implant manufacturers and types from anterior-posterior X-ray images.
• To assess the accuracy and efficiency of the ML model in classifying different TSA implant designs.

Main Methods:

• A convolutional neural network (CNN) model was trained and evaluated on 696 shoulder radiographs.
• The dataset was split into 70% for training and 30% for evaluation.
• The model was designed to classify implants from anterior-posterior X-ray images.

Main Results:

• The ML model achieved an overall accuracy of 93.9% in identifying 10 different implant types (4 rTSA, 6 anatomical TSA).
• Positive predictive value, sensitivity, and F-1 scores were all 94%.
• The average identification time per implant was a rapid 0.110 seconds.

Conclusions:

• This study demonstrates the feasibility of using ML for automated shoulder implant identification from X-rays.
• The developed model shows potential for assisting in preoperative planning for shoulder arthroplasty.
• ML-assisted identification can enhance cost-efficiency and potentially improve patient outcomes in shoulder surgery.