文献速递胃肠外科围术期管理与 ERAS

AI辅助腹腔镜胃肠手术周围神经识别:组织病理学验证

Histopathological proof-of-concept validation of AI-assisted peripheral neural track recognition during laparoscopic gastrointestinal surgery.
PubMed 文献 · Surg Endosc · 2026-09-25

研究首次为AI系统在微创胃肠手术中的解剖识别提供组织病理学验证,但活检刻意局限于标本内微小周围神经,排除主要自主神经和神经丛。

要点

  • AI识别的微小周围神经结构病理证实率37.5%,属高难度验证模型。
  • AI应作为手术导航辅助工具,不能替代外科医生的解剖判断。
  • 需进一步研究提升神经识别准确率并评估临床影响。

核心结论

  • AI辅助周围神经轨道识别在腹腔镜胃肠手术中可行,但病理证实率有限。
  • AI应作为辅助而非替代外科专业知识,需结合解剖位置和情境解读信号。

研究设计与样本

  • 前瞻性观察研究,纳入23例腹腔镜胃肠手术患者。
  • 对EUREKA®系统识别的疑似神经组织进行靶向活检,共32份样本行组织病理学评估。
  • 活检刻意限制在标本内微小周围神经,排除主要自主神经和神经丛。

主要发现

  • 手术类型包括8例右半结肠切除伴完整结肠系膜切除(34.8%)、7例左半结肠切除(30.4%)等。
  • 研究首次对AI系统在微创胃肠手术中的解剖识别进行组织病理学验证。

临床怎么用

  • AI可辅助识别解剖「轨道」以支持保留神经策略,但外科医生须根据解剖位置和情境解读信号。
  • AI不应替代外科专业知识,仅作为辅助工具。

局限性

  • 研究刻意评估最具挑战性的微小周围神经亚群,而非主要自主神经和神经丛,结果不能外推至后者的识别。
  • 样本量小(23例患者、32份活检),需进一步研究提升神经识别并评估临床影响。

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: Nerve injury remains a major cause of functional impairment after gastrointestinal surgery. The EUREKA® artificial intelligence (AI) system enabled real-time intraoperative recognition of anatomical structures including nerves. However, histopathological validation of AI-predicted nerve structures remains lacking. The aim is to provide histopathological validation of AI-assisted peripheral neural track recognition during gastrointestinal surgery. METHODS: A prospective observational study was conducted in 23 patients undergoing laparoscopic gastrointestinal procedures. Targeted biopsies were selectively obtained from structures identified by the EUREKA® system as potential nerve tissue within the extracted surgical specimen. All samples underwent histopathological evaluation. Biopsies were intentionally restricted to peripheral nerves contained within the specimen, representing a challenging validation model, excluding major autonomic nerves and plexuses. RESULTS: The surgical procedures included eight right hemicolectomies with complete mesocolic excision (34.8%), seven left hemicolectomies (30.4%), three conventional right hemicolectomies (13%), two anterior resections of the rectum (8.7%), one splenic flexure resection (4.3%), one subtotal gastrectomy (4.3%), and one total gastrectomy (4.3%). Histopathological analysis of 32 targeted biopsies confirmed peripheral nerve tissue in twelve biopsies (37.5%) obtained from very small peripheral neural structures. CONCLUSIONS: This is the first histopathological validation of anatomical recognition performed by an AI system during minimally invasive gastrointestinal surgery. Although histopathological confirmation was achieved in more than one-third of targeted biopsies obtained from very small peripheral neural structures, the study was intentionally designed to evaluate the most challenging subset of neural anatomy rather than major autonomic nerves and plexuses. These findings support the feasibility of AI-assisted surgical navigation for identifying potential anatomical "tracks" during dissection. However, AI should be considered an adjunct rather than a substitute for surgical expertise, as the surgeon must interpret these signals according to anatomical location and context to support nerve-preserving strategies. Further studies are required to improve nerve recognition and evaluate the clinical impact of AI-guided nerve recognition.

来源:Surg Endosc,查看原文。版权归原作者所有。

引用

  1. Morales-Conde S, Guarnieri E, Valdes-Hernandez J, et al. Histopathological proof-of-concept validation of AI-assisted peripheral neural track recognition during laparoscopic gastrointestinal surgery. Surg Endosc. 2026. PMID: 42789066. doi:10.1007/s00464-026-13422-z

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。