Artificial intelligence-based diagnosis of non-displaced femoral neck fractures shows excellent sensitivity and specificity and reduces the need for computed tomography scans in emergency rooms.

Purpose: The purpose of this study was to evaluate the diagnostic performance of an artificial intelligence (AI)-based system for the detection of non-displaced femoral neck fractures (FNFs) on plain pelvic radiographs and to assess its potential impact on confirmatory computed tomography (CT) utilization. Methods: In this retrospective case-control single-centre study, 394 anteroposterior (AP) pelvic radiographs were analysed, including 197 patients with clinically confirmed non-displaced FNFs (Garden I-II) and 197 control radiographs without fracture. Radiological reports and AI outputs were categorized as positive, negative or doubtful. Doubtful findings were analysed separately. Diagnostic accuracy metrics were calculated, and paired comparisons between radiologists and AI were performed using McNemar's test. A secondary analysis evaluated the theoretical potential of AI to reduce confirmatory CT imaging. Results: The AI software correctly identified 189 of 197 fractures (95.9%) with a false-negative rate of 1.0%, compared with 160 correct detections (81.2%) and a false-negative rate of 7.6% by radiologists. McNemar's test demonstrated significantly fewer missed fractures by the AI software ( Conclusion: Combined radiologist and AI-assisted assessment in the evaluation of non-displaced FNFs on plain AP pelvic radiographs improves the likelihood of detection and accurate diagnosis. The combined assessment yields the lowest miss rate and may reduce the need for confirmatory CT imaging without compromising patient safety, particularly in emergency settings and during off-hours. Level of Evidence: Level III.