AI Tool Improves Intracranial Aneurysm Detection Yield

ARTIFICIAL intelligence aided intracranial aneurysm detection on routine brain CT angiography across adult hospital patient groups. In a prospective shadow-mode evaluation of 3,856 consecutive head CT examinations, an algorithm operated silently alongside practicing radiologists. Adjudication of discordant reads revealed that the software demonstrated a sensitivity of 84.6% compared to 71.8% for radiologists alone, maintaining a specificity above 98%. The platform surfaced 55 true-positive vascular anomalies that human interpreters initially overlooked, yielding an incremental detection ratio of 1.83 and a relative enhanced detection rate of 39.0%.

These findings indicate that integrating machine learning into daily imaging workflows substantially elevates intracranial aneurysm detection without compromising diagnostic precision. The overall gain-to-pain ratio reached 1.20, confirming that incremental true-positive findings outweighed false alerts across the entire health system.

Divergent Operational Performance Across Clinical Settings

Operational efficacy varied significantly depending on the practice setting. Inpatient cohorts derived the greatest clinical value, recording a relative enhanced detection rate of 78.3%, a gain-to-pain ratio of 2.57, and an efficient number-needed-to-examine of 28.9 scans per additional finding. Emergency department examinations maintained a balanced gain-to-pain ratio of 1.00 with an enhanced detection rate of 37.1%. Conversely, outpatient imaging exhibited an unfavorable gain-to-pain ratio of 0.67 and required 130.3 scans to identify one incremental aneurysm, as false alerts exceeded true detections.

This pronounced divergence highlights how underlying disease prevalence, case urgency, and radiologist cognitive workloads shape algorithmic performance. Rather than deploying identical automated screening tools uniformly, healthcare networks should calibrate alert thresholds or tailor implementation specifically to acute care populations.

Complementary Strengths and Downstream Clinical Impact

The study established a complementary diagnostic profile between human specialists and machine assistance. The algorithm preferentially identified diminutive lesions measuring under 3 mm, which accounted for 56.4% of its unique detections. In contrast, radiologists more frequently captured larger aneurysms between 3 and 5 mm as well as lesions within the vertebrobasilar circulation. Subsequent 1-year follow-up of the 55 computer-detected aneurysms led directly to 40 clinic consultations, 40 additional imaging studies, 12 diagnostic angiograms, and 2 successful preventative surgical interventions. Ultimately, collaborative models that pair computer vision with specialist oversight offer the safest strategy to optimize intracranial aneurysm detection and protect patients from sudden subarachnoid hemorrhage.

Reference

Goldberg-Stein S et al. Prospective Shadow-Mode Evaluation of an Artificial Intelligence Tool for Intracranial Aneurysm Detection on CT Angiography: Incremental Yield and Operational Impact. J Am Coll Radiol. 2026;DOI:10.1016/j.jacr.2026.07.018.

Featured Image: AnnaStills on Adobe Stock.