14491
clinical study

False Positive AI Detection of Intracranial Hemorrhage on Emergency Department Head CT: Incidence and Imaging Patterns


Materials & Methods

This retrospective study reviewed all noncontrast head CT examinations performed across a multi-institution Emergency Department (ED) network over a 12-month period that were flagged as positive for intracranial hemorrhage (ICH) by a commercial AI platform. Scans performed overnight (10 p.m.–7:30 a.m.), during which radiologists lacked AI access, were excluded. Each flagged study was classified as a true or false positive, and the specific anatomic findings responsible for false positive detections were documented and categorized.

Results

Out of 14,707 ED CT head studies performed in 2024, 632 (4.3%) received AI-positive notifications. Among these, 243 (38%) were false positives, conferring a positive predictive value of 72.2%. Commonly misclassified hyperdense structures included physiologic calcifications (choroid plexus, falx, tentorium), postoperative changes, cortical/dural calcifications, non-hemorrhagic neoplasms (e.g., meningiomas) and image artifacts such as beam hardening or motion. Heat maps and widget-based review were instrumental in identifying the falsely flagged regions.

Conclusions

AI-generated ICH alerts demonstrate substantial false positive rates, frequently due to benign hyperdensities or artifacts. Understanding common causes of misclassification enables radiologists to triage alerts more efficiently and ensures timely identification of true ICH cases. Enhanced user familiarity with AI tools may mitigate unnecessary workflow interruptions and support more accurate ED imaging interpretation.

Disclaimer: This links to a third-party website, which may require credentials to review.

Explore the Latest AI Insights, Trends and Research