Artificial Intelligence for Surgical Scene Understanding: A Systematic Review and Reporting Quality Meta-Analysis

Area of Science:

• * Computational methods for analyzing surgical video data.
• * Artificial Intelligence (AI) applications in medical imaging.

Background:

• * Surgical Scene Understanding (SSU) utilizes AI to interpret visual data from surgeries, like laparoscopic videos.
• * Despite reported AI capabilities for identifying surgical elements, clinical adoption remains low.
• * Potential for real-time AI-driven support in operating rooms is recognized but not yet realized.

Purpose of the Study:

• * To systematically review and analyze the current state of computational SSU.
• * To identify research gaps in data curation, model design, validation, and clinical applicability.
• * To assess progress toward real-world implementation of SSU.

Main Methods:

• * Systematic review and meta-analysis of 188 studies on intraoperative minimally invasive abdominal surgery data.
• * Inclusion criteria focused on computational SSU methods, trainable models, formal validation, and performance metrics.
• * Analysis covered data characteristics, model development, validation strategies, and clinical translation efforts.

Main Results:

• * Most studies used small, single-center datasets, often from laparoscopic cholecystectomies, lacking diversity and metadata.
• * Research predominantly descriptive, with insufficient reporting on clinical relevance, limitations, code availability, and model uncertainty.
• * Validation methods were often basic, lacking external testing and clinical expert involvement; only 11 studies addressed clinical translation.

Conclusions:

• * Significant research gaps hinder the clinical implementation of SSU.
• * There is a critical need for diverse, multi-institutional datasets and robust validation protocols.
• * Clinically driven development is essential to realize the full potential of SSU in surgical practice.