Accuracy and reliability of artificial intelligence-based automated measurement of hallux valgus-related angles on plain foot radiographs: a systematic review
Year: 2026
Type: article
Abstract: Background This systematic review aims to evaluate the accuracy, reliability, and clinical utility of deep learning algorithms for automated hallux valgus measurements on weight-bearing radiographs. Methods We searched Ovid MEDLINE, Embase, Web of Science, and Cochrane CENTRAL from inception to June 2026 for studies evaluating automated hallux valgus angle (HVA) and intermetatarsal angle (IMA) measurements by deep learning models. Reported measurement errors and reliability estimates were summarized descriptively. Risk of bias and applicability were assessed using an adapted QUADAS-2 framework. Results From 1080 initial records, 14 reports met the eligibility criteria. MAE was reported for HVA in 11 reports and IMA in 10, including one smartphone-based evaluation; six reports provided RMSE for both angles. For primary systems evaluated directly on radiographs, reported MAEs ranged from 0.69° to 4.14° for HVA and from 0.57° to 3.22° for IMA. Corresponding RMSEs ranged from 0.87° to 5.22° and from 0.92° to 3.76°, respectively. Seven reports provided numerical ICCs, which generally indicated high agreement, although differences in ICC definitions and reference measurements limited comparability. Differences in validation datasets, potentially overlapping cohorts, and insufficient reporting of metric-specific uncertainty precluded reliable pooling. Low average errors did not establish that all individual measurements were within clinically acceptable limits. Conclusion Deep learning algorithms showed promising accuracy and agreement with expert measurements of HVA and IMA on plain foot radiographs. However, methodological differences and limitations in the available validation evidence warrant cautious interpretation. Further independent validation, including assessment of individual measurement errors and performance in clinically challenging cases, is needed to establish their suitability for routine clinical use.
Language: English
Subfield: Health Informatics
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $3,090
OpenAlex ID: W7225266553
FWCI:
Cites:
Cited by:
Related to: