用结构化检查提升手术安全评估的准确与可审计性。
Sum-of-Checks: Structured Reasoning for Surgical Safety with Large Vision-Language Models
- 将手术判断拆解为专家定义的视觉验证条目，逐项评估
- 相比基线方法，帧级平均精度提升12%至14%
- 适合医疗AI研发者和手术辅助系统设计者参考
精准评估腹腔镜胆囊切除术中的关键安全视图（CVS）对预防胆管损伤至关重要，该并发症关联高发病率和死亡率。尽管大型视觉语言模型（LVLMs）具备灵活推理能力，但在安全敏感的外科任务中其预测难以审计且不可靠。本文提出Sum-of-Checks框架，将每个CVS标准分解为专家定义的推理检查项，反映临床相关的视觉证据。给定内窥镜图像帧，LVLM对每项检查进行二值判断并提供理由，准则级得分通过固定加权聚合检查结果计算。在Endoscapes2023基准上，使用三种前沿LVLMs进行评估，对比直接提示、思维链及子问题分解方法，均含少量示例。结果显示，Sum-of-Checks在所有三个模型和标准下，相较最佳基线平均帧级平均精度提升12%–14%。个体检查分析表明，LVLM在观察性检查（如可视性、器械遮挡）上表现可靠，但在决定性解剖证据上存在显著波动。结论指出，将证据获取与决策分离的结构化推理能显著提升基于LVLM的CVS评估的准确性与透明度，对构建可审计的外科AI系统具有关键意义。代码已公开于https://github.com/BrachioLab/SumOfChecks。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。