Artificial Intelligence-Assisted Surgical Scene Recognition: A Comparative Study Among Health Care Professionals

Area of Science:

• Neurosurgery
• Artificial Intelligence
• Medical Imaging Analysis

Background:

• Modern microscopic surgery generates vast amounts of operative video data.
• Computer vision, a field within AI, enables automated analysis of surgical videos.
• AI integration in healthcare promises benefits for clinicians, systems, and patients.

Purpose of the Study:

• To compare the diagnostic performance of a deep-learning model (MACSSwin-T) against healthcare professionals in detecting cerebral aneurysms.
• To evaluate the impact of AI assistance on the neurosurgical team's ability to detect cerebral aneurysms.

Main Methods:

• A cross-sectional comparative study involving neurosurgeons, anaesthetists, and OR nurses reviewing operative video frames.
• Participants classified frames as containing or not containing aneurysms, with and without AI assistance.
• Accuracy was calculated for human-only, AI-only, and AI-assisted human detection groups.

Main Results:

• Healthcare professionals achieved 70% accuracy without AI and 78% with AI assistance (OR 1.77, P<0.001).
• Neurosurgical attending physicians showed the most significant improvement, increasing accuracy from 77% to 92% with AI (OR 4.24, P=0.003).
• AI-assisted performance surpassed both human-only and AI-only detection.

Conclusions:

• AI-assisted human performance in cerebral aneurysm detection exceeded that of humans or AI working independently.
• Accuracy improvements were observed across all healthcare professionals, irrespective of subspecialty or experience level.
• AI is a crucial tool for surgical practice, benefiting clinicians across the entire experience spectrum.