Research Information System
Journal of Clinical Medicine, vol.15, no.6, 2026 (SCI-Expanded, Scopus)
Background: Artificial intelligence (AI) is transitioning from proof-of-concept prototypes to clinically utilized tools in orthopedics. The key translational question is whether AI will replace surgeons or, more realistically, augment human expertise. Methods: A structured narrative review was conducted using PubMed/MEDLINE, Web of Science, and Google Scholar (completed 31 January 2026). Peer-reviewed English-language studies that utilized AI for orthopedic clinical problems were eligible. To synthesize the 73 included papers without forced quantitative pooling, evidence was qualitatively charted and organized using a four-axis framework: clinical task, data modality, validation maturity, and intended user/setting. Results: The evidence base was dominated by retrospective, imaging-centered AI studies (predominantly LOE III). Radiograph-based fracture detection and automated measurements were frequently reported to achieve high discrimination, though performance degraded in complex or “edge” cases. Predictive models for arthroplasty and spine outcomes demonstrated variable actionability and inconsistent reporting of calibration. Common translational barriers across subspecialties included limited external validation, dataset shift, and a scarcity of prospective impact studies. Conclusions: Current evidence supports an augmentation paradigm rather than a replacement paradigm. AI acts as a “co-surgeon,” improving triage and standardizing quantification. However, safe clinical translation requires representative external validation, rigorous failure analysis, and human-in-the-loop workflows where surgeons retain ultimate accountability.