Artificial Intelligence-Driven Decision-Making for Knee Joint Manipulation Following Primary Total Knee Arthroplasty

Area of Science:

• Orthopedic Surgery
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Postoperative stiffness is a common complication following total knee arthroplasty (TKA).
• Manipulation under anesthesia (MUA) is frequently used to treat TKA stiffness.
• Optimal MUA timing, motion thresholds, and expected outcomes are not well-defined due to limited evidence.

Purpose of the Study:

• To evaluate an artificial intelligence (AI) platform's ability to create a data-driven decision framework for MUA after primary TKA.
• To define appropriate MUA timing and predict range of motion (ROM) gains using AI modeling.

Main Methods:

• A systematic literature review was conducted on ROM trajectory and manipulation following primary TKA.
• Data from the literature review were used to train an AI model.
• AI scenarios modeled variables like time from surgery, knee flexion, extension deficit, preoperative ROM, and BMI to determine MUA appropriateness and expected ROM improvements.

Main Results:

• AI model indicates MUA is appropriate for patients with knee flexion < 80 degrees and/or extension deficits > 20 degrees at six weeks post-TKA.
• Predicted mean ROM gains with MUA are 26 degrees in flexion and 3 degrees in extension.
• Delaying MUA beyond 90 days reduced expected flexion gains; combining MUA with arthroscopic lysis of adhesions is advised for persistent stiffness beyond three months.

Conclusions:

• An AI-generated decision framework for MUA aligns with current consensus on timing and patient selection.
• The AI tool should supplement, not replace, clinical judgment due to evidence limitations and the complexity of postoperative stiffness.