Artificial Intelligence Templating for Total Knee Arthroplasty: Using Standard Preoperative Knee Radiographs

Area of Science:

• Orthopedic Surgery
• Medical Imaging
• Artificial Intelligence

Background:

• Preoperative planning in total knee arthroplasty (TKA) is crucial for optimal outcomes.
• Accurate templating of implant sizes is a key component of TKA preoperative planning.
• Traditional methods rely on 2D radiographs, which have limitations in predicting 3D bone geometry.

Purpose of the Study:

• To evaluate the accuracy of a novel artificial intelligence (AI) algorithm in predicting TKA implant sizes.
• To assess the learning capabilities and improvement in accuracy of the AI algorithm over time.
• To determine the potential of AI-driven 3D bone modeling for TKA preoperative planning.

Main Methods:

• Retrospective analysis of 108 primary TKA cases.
• Utilized a novel AI algorithm to convert 2D knee radiographs into 3D bone models.
• Compared AI-predicted implant sizes with actual implanted component sizes for accuracy assessment.
• Evaluated AI learning by comparing accuracy in the first 50 patients versus the last 51 patients.

Main Results:

• The AI algorithm achieved exact prediction accuracy of 40% for femoral and 41.5% for tibial components.
• Allowing for a ±1 size difference, accuracy improved to 90% for femoral and 87% for tibial components.
• The AI algorithm demonstrated significant improvement in accuracy over time for tibial components (p=0.0342) and a trend for femoral components (p=0.1719).

Conclusions:

• The AI-based algorithm accurately predicts TKA implant sizes from 2D radiographs.
• The algorithm shows enhanced accuracy over time, indicating a learning capability.
• This AI tool holds promise as a valuable aid for preoperative planning in total knee arthroplasty.