Accuracy evaluation of artificial intelligence-based mobile applications for assessing traumatic injury severity through photographic analysis: Simulation study

Abstract

Introduction: Rapid and accurate assessment of traumatic injury severity is crucial for effective triage and timely intervention in emergency medicine. Mobile applications based on artificial intelligence (AI) offer an objective assessment of injury severity through photographic analysis; however, their accuracy has not been sufficiently explored. Objectives: To evaluate the accuracy of available AI-based mobile applications for assessing traumatic injury severity in simulated conditions. Methods: This simulation study tested five mobile applications (DermaScore AI, SkinVision, Tissue Analytics, WoundCheck AI, and BurnCare App). A total of 200 simulated images of traumatic injuries, classified by the Abbreviated Injury Scale (AIS) as mild, moderate, severe, and critical, were analyzed by each application. Accuracy, sensitivity, specificity, and ROC analysis were evaluated, along with the impact of photo resolution on app performance. Results: DermaScore AI achieved the highest overall accuracy (89%), sensitivity (92%), and ROC-AUC value (0.91). The lowest accuracy was recorded by BurnCare App (74%). Higher photo resolution (above 12 MP) significantly improved the accuracy of all tested apps (p = 0.0014). Conclusion: AI-based mobile applications can reliably assess traumatic injury severity from photographic analysis, but their performance significantly varies depending on technical and algorithmic factors. Additional clinical research is required to validate these findings in real-world settings.

Keywords

• mobile applications
• traumatic injuries
• artificial intelligence
• emergency medicine
• simulation study