Improving Radiographic Fracture Recognition Performance and Efficiency Using Artificial Intelligence

• DOI: 10.1148/radiol.210937
• Publication Date: 2021-12-21T15:00:01Z
• Authors: Ali Guermazi; Chadi Tannoury; Andrew J. Kompel; Akira M. Murakami; Alexis Ducarouge; André Gillibert; Xinning Li; Antoine Tournier; Youmna Lahoud; Mohamed Jarraya; Elise Lacave; Hamza Rahimi; Aloïs Pourchot; Robert L. Parisien; Alexander C. Merritt; Douglas Comeau; Nor-Eddine Regnard; Daichi Hayashi
• Keywords: Gold standard (test)

Background Missed fractures are a common cause of diagnostic discrepancy between initial radiographic interpretation and the final read by board-certified radiologists. Purpose To assess effect assistance artificial intelligence (AI) on performances physicians for radiographs. Materials Methods This retrospective study used multi-reader, multi-case methodology based an external multicenter data set 480 examinations with at least 60 per body region (foot ankle, knee leg, hip pelvis, hand wrist, elbow arm, shoulder clavicle, rib cage, thoracolumbar spine) July 2020 January 2021. Fracture prevalence was 50%. The ground truth determined two musculoskeletal radiologists, discrepancies solved third. Twenty-four readers (radiologists, orthopedists, emergency physicians, physician assistants, rheumatologists, family physicians) were presented whole validation (n = 480), without AI assistance, 1-month minimum washout period. primary analysis had to demonstrate superiority sensitivity patient noninferiority specificity -3% margin aid. Stand-alone performance also assessed using receiver operating characteristic curves. Results A total patients included (mean age, 59 years ± 16 [standard deviation]; 327 women). 10.4% higher (95% CI: 6.9, 13.9; P < .001 superiority) aid (4331 5760 readings, 75.2%) than (3732 64.8%). (5504 95.6%) noninferior that (5217 90.6%), difference +5.0% +2.0, +8.0; noninferiority). shortened average reading time 6.3 seconds examination -12.5, -0.1; .046). gain significant in all regions (+8.0% +16.2%; .05) but clavicle spine (+4.2% +2.6%; .12 .52). Conclusion improved may even improve fracture detection radiologists nonradiologists, lengthening time. Published under CC BY 4.0 license. Online supplemental material is available this article. See editorial Link Pedoia issue.