Sum-of-Checks: Structured Reasoning for Surgical Safety with Large Vision-Language Models

TL;DR

Sum-of-Checks enhances surgical safety assessment by decomposing complex visual reasoning into expert-defined checks, improving accuracy and transparency of large vision-language models in critical laparoscopic procedures.

Contribution

It introduces a structured framework that decomposes surgical reasoning into verifiable checks, improving reliability and auditability of AI in surgical safety assessment.

Findings

Sum-of-Checks improves average frame-level mean average precision by 12-14%.

LVLMs are reliable on observational checks but variable on anatomical evidence.

Explicitly separating evidence from decision-making enhances AI transparency in surgery.

Abstract

Purpose: Accurate assessment of the Critical View of Safety (CVS) during laparoscopic cholecystectomy is essential to prevent bile duct injury, a complication associated with significant morbidity and mortality. While large vision-language models (LVLMs) offer flexible reasoning, their predictions remain difficult to audit and unreliable on safety-critical surgical tasks. Methods: We introduce Sum-of-Checks, a framework that decomposes each CVS criterion into expert-defined reasoning checks reflecting clinically relevant visual evidence. Given a laparoscopic frame, an LVLM evaluates each check, producing a binary judgment and justification. Criterion-level scores are computed via fixed, weighted aggregation of check outcomes. We evaluate on the Endoscapes2023 benchmark using three frontier LVLMs, comparing against direct prompting, chain-of-thought, and sub-question decomposition, each with and without few-shot examples. Results: Sum-of-Checks improves average frame-level mean average precision by 12--14% relative to the best baseline across all three models and criteria. Analysis of individual checks reveals that LVLMs are reliable on observational checks (e.g., visibility, tool obstruction) but show substantial variability on decision-critical anatomical evidence. Conclusion: Structuring surgical reasoning into expert-aligned verification checks improves both accuracy and transparency of LVLM-based CVS assessment, demonstrating that explicitly separating evidence elicitation from decision-making is critical for reliable and auditable surgical AI systems. Code is available at https://github.com/BrachioLab/SumOfChecks.