AI May Have Value Assisting Clinician Detection Accuracy in Cerebral Aneurysm

Williams SC, Zhou J, Muirhead WR, Khan D, et al. Artificial Intelligence–Assisted Surgical Scene Recognition: A Comparative Study Among Health Care Professionals. Ann Surg. May 2026.

This study set out to evaluate how a deep-learning system (MACSSwin-T) performs in identifying cerebral aneurysms in surgical video compared with healthcare professionals and examined whether clinicians’ detection accuracy improves when supported by artificial intelligence (AI).

This cross-sectional comparative study involved neurosurgeons, anesthetists, and OR nurses with varying levels of training and experience. Participants reviewed still images from aneurysm clipping procedures and classified each frame as either containing an aneurysm or not. The same frames were then analyzed by the AI model. In the second phase, clinicians repeated the task with AI support. Detection accuracy was compared across three groups: clinicians alone, AI alone, and clinicians assisted by AI.

A total of 5,154 frame evaluations were collected from 338 health care professionals. Without AI support, clinicians correctly identified aneurysms in 70% of cases; this increased to 78% with AI assistance (odds ratio 1.77, P < 0.001). The greatest improvement was observed among attending neurosurgeons, whose accuracy rose from 77% to 92% when aided by AI (odds ratio 4.24, P = 0.003).

Performance was highest when clinicians used AI, exceeding both human-only and AI-only results. Accuracy improved across all specialties and experience levels, with particularly strong gains among the most experienced clinicians. 

These findings challenge the common assumption that AI primarily benefits less experienced practitioners, underscoring its potential value across all levels of surgical expertise and its importance in modern surgical practice.