Using artificial intelligence, computer vision systems can analyze videos of robotic-assisted surgeries and generate post-surgical operative reports with greater accuracy than surgeons' own documentation, a new study suggests.

Operative reports—essential for clinical communication, surgical billing, quality benchmarking, research, and regulatory compliance—are often tedious to write and prone to inaccuracies, researchers reported in the Journal of the American College of Surgeons.

For this study, researchers trained an AI model to recognize specific surgical steps in videos of robotic-assisted prostatectomies. Each operative step—such as lymph node dissection, vessel ligation, or urethral transection—was paired with pre-written descriptive text. The AI then compiled these into a structured operative report.

When tested on videos from 158 cases, 53% of surgeon-generated reports contained discrepancies, compared with 29% of AI-generated reports. Expert reviewers identified clinically significant discrepancies in 27% of surgeon reports versus 13% of AI reports.

The researchers noted that with further refinement, this technology could reduce documentation burden, enhance accuracy, improve transparency, and minimize subjectivity in surgical documentation.