Background

Artificial Intelligence (AI) has become increasingly integrated into both daily life and professional practices, particularly as a decision-support tool for radiological diagnosis in emergency departments.

Objective

To evaluate the diagnostic performance of an AI-assisted radiographic software (Deep Unity Gleamer BoneView) for the detection of metacarpal and phalangeal fractures and dislocations on hand radiographs.

Methods and Materials

This was a retrospective, diagnostic, monocentric study conducted in an emergency department. The study population included all patients admitted for hand trauma who underwent hand radiography.

Each radiograph was independently reviewed by two senior hand surgeons to determine the presence or absence of fractures or dislocations. In cases of disagreement, a consensus gold standard (GS) was established after re-evaluation, incorporating clinical data.

The diagnostic performance of the AI software was assessed against the GS and expressed in terms of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), with 95% confidence intervals (CI).

Between December 2022 and January 2024, a total of 1,915 hand radiographic records were analyzed, corresponding to 1,892 patients and 4,738 individual X-ray images.
The mean patient age was 40.6 years (range: 16–99).

According to the GS:

• 457 radiographs showed fractures/dislocations
• 1,458 radiographs showed no lesions

The AI system identified:

• 608 fractures/dislocations
• 1,307 radiographs without lesions

Results

The AI achieved:

• Sensitivity: 97.6% (95% CI: 95.7–98.7)
• Specificity: 88.9% (95% CI: 87.2–90.4)
• PPV: 73.4% (95% CI: 69.8–76.9)
• NPV: 99.2% (95% CI: 98.7–99.7)

Upon re-evaluation, 11 false-negative cases were identified.

Overall, the AI demonstrated excellent sensitivity and strong diagnostic performance in detecting metacarpal and phalangeal fractures/dislocations, consistent with previously reported results for carpal and distal radius injuries.

Conclusions

AI-assisted radiographic software shows excellent diagnostic performance, particularly high sensitivity, for the detection of metacarpal and phalangeal fractures and dislocations in adults. While AI represents a valuable support tool in emergency settings, it should complement—rather than replace—clinical expertise and comprehensive patient evaluation.

Note: The full article is available in French only.

‍