Artificial intelligence-based versus conventional preoperative templating for tibial and femoral component size estimation in total knee arthroplasty: a systematic review and meta-analysis.

BACKGROUND: Preoperative templating supports total knee arthroplasty (TKA) planning, and AI-assisted tools are increasingly used to improve component-size estimation. However, the magnitude of improvement in exact femoral and tibial component-size prediction compared with conventional templating has not been quantitatively synthesized. We aimed to compare the accuracy of AI-assisted preoperative templating with conventional techniques in predicting exact implanted prosthesis size. METHODS: PubMed (MEDLINE), Scopus, Web of Science, and the Cochrane Library were searched from inception to November 2025. Eligible studies included patients undergoing primary TKA in whom AI-assisted preoperative templating predicted exact femoral and/or tibial component size and was compared with conventional templating. Risk of bias was assessed using RoB 2 and ROBINS-I. RESULTS: Eight studies were included. In implanted-size-restricted analyses, AI-assisted templating did not significantly improve exact femoral component-size prediction (risk ratio (RR) 1.38; 95% confidence interval (CI) 0.92-2.08; P = 0.09; 95% prediction interval 0.63-3.04), whereas tibial prediction was borderline significant (RR 1.38; 95% CI 1.01-1.89; P = 0.05; 95% prediction interval 0.76-2.49). Computed tomography (CT)-based AI-assisted planning was associated with higher exact femoral and tibial prediction accuracy than conventional radiographic templating, while radiograph-based AI analyses were inconclusive. Short-term functional scores at 3 months after surgery showed no significant between-group differences. CONCLUSION: CT-based AI-assisted templating may improve exact component-size prediction in primary TKA. However, prediction intervals crossed the null, indicating uncertainty regarding the consistency and magnitude of benefit. The apparent advantage may reflect combined effects of three-dimensional imaging and AI-assisted planning rather than AI alone. Evidence regarding radiograph-based AI planning and postoperative benefit remains inconclusive.