A German prospective study evaluated BoneView (Gleamer), Rayvolve (AZmed), and RBfracture (Radiobotics), AI tools for the detection of fractures, dislocations, and joint effusions, using radiographs of 1,037 adult participants across 22 anatomical regions. The AI tools were integrated into routine clinical workflow at a tertiary center and processed 2,926 radiographs from emergency, outpatient, and inpatient departments. AI outputs were compared to radiologist reports, with CT confirmation where available.
For all fractures, AZmed (AUC 84.88%, sensitivity 79.48%) and Gleamer (AUC 83.95%, sensitivity 75.57%) outperformed Radiobotics (AUC 77.24%, sensitivity 60.91%). For acute fractures, performance was comparable across tools (AUC ~85–88%, accuracy ~91%). AZmed was best for multiple fractures (AUC 73.40%) and dislocation detection (AUC 61.85%), while Gleamer and Radiobotics showed better performance for effusions (AUC 69.59% and 73.63%, respectively).
While AI tools performed well for routine fracture detection, they underperformed in complex scenarios such as dislocations, multiple fractures, and cases involving metallic implants. Performance also varied by region, with reduced accuracy in anatomically complex areas like the foot. As the authors state, “none of the evaluated algorithms are robust enough for stand-alone clinical use,” and all require radiologist oversight. They emphasize the need for broader validation and more diverse training data to improve generalizability.
Evaluation of commercial AI algorithms for the detection of fractures, effusions, and dislocations on real-world clinical data: A prospective registry study
Radiography, 2025
Abstract
Purpose
To prospectively evaluate and directly compare the performance of three commercial AI algorithms (Gleamer, AZmed, and Radiobotics) for detecting fractures, dislocations, and joint effusions across multiple anatomical regions in real-world adult clinical radiography.
Material and methods
In this single-center, prospective technical performance evaluation study, we assessed these algorithms on radiographs from adult patients (n = 1037; 2926 radiographs; 22 anatomical regions) at the Technical University of Munich (January–March 2025). Radiologists’ reports served as the reference standard, with CT adjudication when available. Sensitivity, specificity, accuracy, and AUC were calculated; AUCs were compared using Bonferroni-corrected DeLong tests.
Results
Fractures were identified in 29.60 % of patients; 13.69 % had acute fractures and 6.65 % had multiple fractures. For all fractures, Gleamer (AUC 83.95 %, sensitivity 75.57 %, specificity 92.33 %) and AZmed (AUC 84.88 %, sensitivity 79.48 %, specificity 90.27 %) outperformed Radiobotics (AUC 77.24 %, sensitivity 60.91 %, specificity 93.56 %). For acute fractures, AUCs were comparable (range: 84.81–87.78 %). For multiple fractures, performance was limited (AUCs 64.17–73.40 %). AZmed had higher AUC for dislocation (61.85 % vs. 54.48 % for Gleamer), while Gleamer and Radiobotics outperformed AZmed for effusion (AUC 69.59 % and 73.63 % vs. 57.99 %). No algorithm exceeded 91 % accuracy for acute fractures.
Conclusion
In this real-world, single-center study, commercial AI algorithms showed moderate to high performance for straightforward fracture detection but limited accuracy for complex scenarios such as multiple fractures and dislocations.
