Artificial Intelligence in Surgery: A Systematic Review of Use and Validation

Area of Science:

• Medical Informatics
• Surgical Technology
• Artificial Intelligence in Healthcare

Background:

• Artificial Intelligence (AI) is increasingly used in surgical settings, yet lacks standardized validation protocols.
• Ensuring the accuracy and patient safety of AI models is critical before widespread clinical adoption.
• Suboptimal validation risks integrating unproven AI, potentially impacting patient outcomes negatively.

Purpose of the Study:

• To review current AI model validation and usage in clinical surgery.
• To propose a novel classification system for AI model validation in surgery.

Main Methods:

• Systematic review of PubMed and Cochrane databases.
• Keywords: "validation", "artificial intelligence", "surgery".
• PRISMA guidelines followed for study selection and data extraction.

Main Results:

• 102 studies (2,837,211 patients) were included from 7627 initial articles.
• Developed the Surgical Validation Score (SURVAS) classification system.
• Only 45% of studies used high-evidence validation methods; 14% provided public datasets. AI primarily used for preoperative risk assessment.

Conclusions:

• AI has significant surgical applications, primarily in preoperative risk assessment and decision-making.
• Current AI validation quality is suboptimal, with limited public data availability.
• Classification systems like SURVAS can aid clinicians in assessing AI model validity prior to clinical use.