Related Experiment Video
Updated: Mar 20, 2026
09:31
Individualized Stem-positioning in Calcar-guided Short-stem Total Hip Arthroplasty
Published on: February 27, 2018
12.6K
Deviation From AI-Predicted Implant Sizes in Total Hip Arthroplasty Is Associated With Increased Complication Risk
Michael P Murphy1, Andrew M Schneider2, Cameron J Killen3
1Department of Orthopaedic Surgery, Anderson Orthopaedic Research Institute, Alexandria, VA, USA.
Arthroplasty Today
|March 19, 2026
Summary
An AI tool accurately predicts total hip arthroplasty implant sizes using patient data, outperforming traditional methods. This AI shows promise for improving surgical planning and outcomes, especially in standard cases.
Area of Science:
- Orthopedic surgery
- Artificial intelligence in medicine
- Biomedical engineering
Background:
- Predicting total hip arthroplasty (THA) implant size using patient characteristics is valuable but lacks proven clinical utility.
- Evaluating the accuracy and generalizability of AI for THA sizing is crucial.
Purpose of the Study:
- To assess an AI-driven tool's accuracy in predicting femoral stem and acetabular cup sizes for primary THA.
- To compare AI predictions with implanted sizes and traditional radiographic templating.
- To investigate the association between AI size prediction accuracy and surgical outcomes.
Main Methods:
- Retrospective review of 2410 primary THAs across two academic hospitals.
- Utilized an AI model predicting implant sizes based on patient demographics (age, sex, height, weight, race/ethnicity).
- Compared AI predictions to actual implanted sizes and radiographic templating, analyzing complications like periprosthetic femur fracture and aseptic loosening.
Main Results:
- AI predictions were within one size of implanted components for 72.0% of stems and 77.7% of cups.
- AI outperformed radiographic templating for both stem and cup size predictions (P < .0001 for stems, P = .036 for cups).
- Closer alignment with AI-predicted stem size correlated with reduced aseptic loosening and periprosthetic femur fracture (P < .0001).
Conclusions:
- The AI tool (Mortho) demonstrated high accuracy and generalizability for primary THA implant sizing, especially in standard cases.
- AI predictions aligned with selections made by arthroplasty-trained surgeons.
- AI-based prediction is a valuable adjunct for THA planning, potentially improving outcomes.
