AI assistance improves rib fracture detection sensitivity and reduces diagnostic time in adults
A systematic review and meta-analysis examined the diagnostic performance of clinicians assisted by artificial intelligence (AI) versus unassisted clinicians for detecting traumatic chest injuries on imaging in adults. The analysis included 20 studies, with 12 suitable for meta-analysis specifically for rib fracture detection. The setting was not reported.
For the primary outcome of diagnostic sensitivity, AI assistance was associated with improvement. The mean difference in sensitivity was 0.12, with absolute numbers of 0.88 for AI-assisted clinicians (CA) versus 0.76 for unassisted clinicians (CU). For the secondary outcome of diagnostic time, AI assistance was associated with a reduction. The mean difference was -99 seconds, with absolute times of 115 seconds for CA versus 214 seconds for CU. P-values or confidence intervals were not reported for these outcomes.
Safety and tolerability data, including adverse events and discontinuations, were not reported. The authors explicitly noted a key limitation: the overall quality of the evidence is poor. They also cautioned against overstating causation, generalizability beyond rib fractures, and clinical utility.
In restrained practice relevance, the analysis suggests AI assistance can be associated with improved diagnostic performance for clinicians in this specific task. However, the poor evidence quality and lack of safety data mean these findings are preliminary. Further research into clinically useful and validated models is required before any definitive conclusions about integration into routine practice can be made.
View Original Abstract ↓
Muscle ultrasonography shows high diagnostic accuracy for fasciculation detection in ALS
A Simple Scan Could Speed Up an ALS Diagnosis
Related in Radiology & Imaging
From Other Specialties
