Diagnostic Performance of Artificial Intelligence in Rib Fracture Detection: Systematic Review and Meta-Analysis

Abstract

Artificial intelligence (AI) is a promising tool for diagnosing rib fractures. To date, only a few studies have quantified its performance. The objective of this systematic review was to assess the accuracy of AI as an independent tool for rib fracture detection on CT scans or radiographs. This was defined as the combination of sensitivity and specificity. PubMed (including MEDLINE and PubMed Central) was systematically reviewed according to the PRISMA statement followed by citation searching among studies up to December 2022. Methods of the analysis and inclusion criteria were prespecified in a protocol and published on PROSPERO (CRD42023479590). Only diagnostic studies of independent AI tools for rib fracture detection on CT scans and X-rays reporting on sensitivity and/or specificity and written in English were included. Twelve studies met these criteria, which included 11,510 rib fractures in total. A quality assessment was performed using an altered version of QUADAS-2. Random-effects meta-analyses were performed on the included data. If specificity was not reported, it was calculated on a set of assumptions. Pooled sensitivity and specificity were 0.85 (95% CI, 0.78–0.92) and 0.96 (95% CI, 0.94–0.97), respectively. None of the included studies used X-rays. Thus, it can be concluded that AI is accurate in detecting rib fractures on CT scans. Overall, these findings seemed quite robust, as can be concluded from the study quality assessment, therefore AI could potentially play a substantial role in the future of radiological diagnostics.

Keywords

• artificial intelligence
• rib fracture
• sensitivity
• specificity
• tomography
• X-ray computed