Artificial intelligence‐based assessment of leg axis parameters shows excellent agreement with human raters: A systematic review and meta‐analysis
Mikhail Salzmann, Hakam Hassan Tarek, Robert Prill, Roland Becker, A. Schreyer, Robert Hable, Marko Ostojic, Nikolai Ramadanov
Abstract
PURPOSE: The aim of this study was to conduct a systematic review and meta-analysis on the reliability and applicability of artificial intelligence (AI)-based analysis of leg axis parameters. We hypothesized that AI-based leg axis measurements would be less time-consuming and as accurate as those performed by human raters. METHODS: The study protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO). PubMed, Epistemonikos, and Web of Science were searched up to 24 February 2024, using a BOOLEAN search strategy. Titles and abstracts of identified records were screened through a stepwise process. Data extraction and quality assessment of the included papers were followed by a frequentist meta-analysis employing a common effect/random effects model with inverse variance and the Sidik-Jonkman heterogeneity estimator. RESULTS: A total of 13 studies encompassing 3192 patients were included in this meta-analysis. All studies compared AI-based leg axis measurements on long-leg radiographs (LLR) with those performed by human raters. The parameters hip knee ankle angle (HKA), mechanical lateral distal femoral angle (mLDFA), mechanical medial proximal tibial angle (mMPTA), and joint-line convergence angle (JLCA) showed excellent agreement between AI and human raters. The AI system was approximately 3 min faster in reading standing long-leg anteroposterior radiographs (LLRs) compared with human raters. CONCLUSION: AI-based assessment of leg axis parameters is an efficient, accurate, and time-saving procedure. The quality of AI-based assessment of the investigated parameters does not appear to be affected by the presence of implants or pathological conditions. LEVEL OF EVIDENCE: Level I.