Neuro / Head & NeckEmergencyAI / InformaticsResearchTrainee
Aidoc AI Falls Short of Radiologists for Acute Cervical Spine Fracture Detection on CT
American Journal of Neuroradiology (AJNR)2w ago
In 2,652 CT exams (4.1% fracture prevalence), Aidoc AI detected acute cervical spine fractures with sensitivity 89.1% vs 94.5% for radiologists; combined sensitivity/specificity favored radiologists (p=0.02). Single-center retrospective study.
- Retrospective single-center study (UW Medicine, Nov 2020–Mar 2021); 2,652 CT exams across four scan types; 110 acute fractures confirmed by consensus radiologist panel as reference standard.
- AI specificity was 99.2% (95% CI 98.7–99.5%) vs radiologist 99.7% (95% CI 99.4–99.9%); 21 AI false positives driven by atherosclerotic calcifications, chronic fractures, beam artifact, osteophytes, and unfused apophysis.
- Key limitation: retrospective, single-institution design with low fracture prevalence (4.1%) limits generalizability; no external or prospective validation of the Aidoc tool in this cohort.
RadPigeon summaries are original and for information only. They are not clinical advice.
