文献速递疝与腹壁外科围术期管理与 ERAS

人工智能在腹壁疝手术中的临床应用与转化准备度

Artificial intelligence in abdominal wall hernia surgery: clinical applications and translational readiness.
PubMed 文献 · Surg Endosc · 2026-09-25

该叙述性综述综合了PubMed/MEDLINE、Web of Science和Scopus从建库至2025年12月31日的文献,并于2026年1月15日补充检索,探讨AI在腹壁疝围手术期中的应用。AI应用集中在术前风险分层与CT体成分分析、术中计算机视觉、术后结局预测与数字随访、注册登记大数据分析四个领域。术前机器学习风险模型和CT肌少症量化最接近临床转化,但证据仍以回顾性为主;临床采纳取决于前瞻性外部验证、患者层面结局获益证明及遵循TRIPOD+AI等报告标准。

要点

  • 术前机器学习风险模型与CT肌少症量化是当前最接近临床转化的AI应用。
  • 术中实时标志识别、增强现实与联邦学习仍处于早期转化阶段。
  • 临床采纳需前瞻性外部验证、患者层面获益证明及遵循TRIPOD+AI等标准。

核心结论

  • 腹壁疝手术因解剖标准化、流程可重复及注册登记和视频资源增长,适合AI整合。
  • 本综述的独特贡献是提出基于标准的临床准备度框架和以指南为锚的转化路线图,而非重新编目现有研究。

研究设计与样本

  • 叙述性综述,检索PubMed/MEDLINE、Web of Science和Scopus从建库至2025年12月31日,并于2026年1月15日补充验证检索。
  • 研究按主题而非系统性地选择,以进行面向临床的综合;两名具有临床和数据科学背景的审阅者独立筛选,全文一致性的Cohen κ=0.82。
  • 报告遵循SANRA(六项,每项0-2分,最高12分;作者自评10/12)。

主要发现

  • AI应用聚集于四个领域:术前风险分层和基于CT的体成分分析;术中计算机视觉用于阶段识别和关键结构检测;术后结局预测和数字随访;基于注册登记的大数据分析。
  • 术前机器学习风险建模和基于CT的肌少症量化最接近临床转化,但证据仍主要为回顾性且经内部验证。
  • 实时术中标志检测、增强现实和联邦学习仍处于较早的转化阶段。

临床怎么用

  • 临床采纳AI工具前应要求前瞻性外部验证和患者层面结局获益的证明。
  • AI研究报告应遵循TRIPOD+AI、CONSORT-AI和SPIRIT-AI标准。
  • 需明确AI作为软件医疗器械的监管路径。

局限性

  • 证据仍以回顾性研究为主,且多为内部验证。
  • 研究按主题而非系统性地选择,可能引入选择偏倚。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

BACKGROUND: Abdominal wall hernia repair is amongst the highest-volume operations in general surgery, yet recurrence, mesh-related morbidity, and surgeon-dependent variability remain unsolved. Artificial intelligence (AI) is increasingly applied across the perioperative continuum of hernia care. METHODS: We performed a narrative synthesis of PubMed/MEDLINE, Web of Science, and Scopus from database inception to 31 December 2025, with a supplementary verification search on 15 January 2026 and eligibility based on the earliest online (Online-First/Epub) date. Studies were selected thematically rather than systematically for a clinically oriented synthesis. Two reviewers with clinical and data-science backgrounds screened independently; full-text agreement was substantial (Cohen κ = 0.82). Reporting followed the Scale for the Assessment of Narrative Review Articles (SANRA; six items scored 0-2, maximum 12; author self-assessment 10/12). KEY FINDINGS: AI applications cluster into four domains: preoperative risk stratification and CT-based body composition analysis; intraoperative computer vision for phase recognition and critical structure detection; postoperative outcome prediction and digital follow-up; and registry-based big data analytics. Preoperative machine learning risk modeling and CT-based sarcopenia quantification are closest to clinical translation, although the evidence remains predominantly retrospective and internally validated; real-time intraoperative landmark detection, augmented reality, and federated learning remain at earlier translational stages. CONCLUSIONS: Hernia surgery is well suited for AI integration because of its standardized anatomy, repeatable workflows, and growing registry and video repositories. The distinct contribution of this review is a criteria-based clinical readiness framework and a guideline-anchored translational roadmap, rather than a re-cataloguing of existing studies. Clinical adoption now depends on prospective external validation, demonstration of patient-level outcome benefit, adherence to TRIPOD+AI, CONSORT-AI, and SPIRIT-AI standards, and clarification of regulatory pathways for AI as a software medical device.

来源:Surg Endosc,查看原文。版权归原作者所有。

引用

  1. Bademkıran E, Bademkıran S. Artificial intelligence in abdominal wall hernia surgery: clinical applications and translational readiness. Surg Endosc. 2026. PMID: 42791408. doi:10.1007/s00464-026-13394-0

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。