Purpose
This research evaluated the diagnostic performance of two AI software programs (BoneView and RBfracture™ version 2.1) in assisting non-specialist radiologists (NSRs) in detecting scaphoid fractures using conventional wrist radiographs (X-rays).
The primary objective of the study was to evaluate the diagnostic performance of two NSRs, with and without support from two AI tools, in detecting scaphoid fractures on plain radiographs.
Key findings
- There was no difference in the discriminative power of RBfracture and BoneView, however, RBfracture demonstrated a higher AUC of 0.84 in Group 1 and 0.65 in Group 2 (compared to 0.83 and 0.62)
- RBfracture exhibited a higher specificity (97.9%) than BoneView (85.1%), potentially reducing unnecessary follow-ups
- BoneView exhibited a higher sensitivity (81.2%) than RBfracture (70%), potentially reducing missed findings
- RBfracture may be more suitable for specialized musculoskeletal radiology settings, where its higher specificity could aid in resolving diagnostic uncertainties among expert radiologists
- AI support significantly increased the diagnostic confidence of non-specialized radiologists in complex cases
Open access
This research is open access
This original research can be read in La Radiologia Medica, published in August 2025.
Funding and conflicts of interests
The authors have no conflicts of interest to declare. There are no financial relationships with the companies that developed the AI software used in this study (Gleamer and Radiobotics).
