Scholarlink

Background <p>Pelvic autonomic nerve preservation (ANP) during total mesorectal excision (TME) is technically demanding. Despite nerve-sparing techniques, urinary dysfunction affects 0–33% and sexual dysfunction 11–75% of patients. Conventional nerve identification relies heavily on surgeon experience and subtle anatomical landmarks. Artificial intelligence (AI)-based computer vision offers standardized, real-time recognition of autonomic nerves and fascial planes independent of individual expertise.</p> Methods <p>This narrative review comprehensively searched PubMed, PubMed Central, Embase, Web of Science, Scopus, and Google Scholar through May 15, 2026, using all relevant keywords. Peer-reviewed studies reporting AI-based neural or fascial plane recognition applicable to pelvic nerve protection during rectal cancer surgery were included. The manuscript is a narrative review; no protocol was registered, as systematic review methodology was not employed.</p> Results <p>The AI Neurorecognition System (AINS) achieved a mean intersection-over-union (mIoU) of 0.75 and 96.09% overall accuracy, reducing nerve recognition time from 25 to 3&#xa0;min. The Eureka navigation system improved trainee recognition in 20.7–90.6% of missed scenarios without increasing complications. Fascial plane models achieved Dice coefficients of 90.4–90.6%. Robotic micro-space dissection demonstrated acceptable oncologic efficacy with reduced dysfunction. The NEUROS multicenter randomized controlled trial demonstrated that intraoperative neuromonitoring (IONM) significantly reduced urinary and sexual dysfunction compared with standard TME at 12 months — the highest-quality evidence currently available for any nerve-preservation adjunct, providing a validation benchmark for future AI trials.</p> Conclusion <p>AI-guided neural recognition has progressed to clinically deployed systems demonstrating high technical accuracy and significant educational benefits. However, no randomized trial has yet demonstrated that AI guidance reduces patient-level functional dysfunction. The transition to standard practice requires multicenter randomized controlled trials measuring validated functional outcomes, head-to-head comparisons with existing nerve-preservation strategies (particularly IONM), rigorous external validation, formal cost-effectiveness analyses, and clear regulatory frameworks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial Intelligence-Guided Pelvic Autonomic Nerve Preservation in Rectal Cancer Surgery: Technical Performance, Clinical Translation, and Critical Re-evaluation of the Field

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Amirhosein Naseri,
  • Mohammad Hossein Antikchi,
  • Abolhasan Alijanpour,
  • Hossein Najdsepas,
  • Hossein Negahban,
  • Alireza Negahi,
  • Mehdi Khosravi-Mashizi,
  • Amirhossein Rahmani,
  • Amirhossein Shahbazi,
  • Ali Massoudi,
  • Kazem Aghili,
  • Hossein Neamatzadeh

摘要

Background

Pelvic autonomic nerve preservation (ANP) during total mesorectal excision (TME) is technically demanding. Despite nerve-sparing techniques, urinary dysfunction affects 0–33% and sexual dysfunction 11–75% of patients. Conventional nerve identification relies heavily on surgeon experience and subtle anatomical landmarks. Artificial intelligence (AI)-based computer vision offers standardized, real-time recognition of autonomic nerves and fascial planes independent of individual expertise.

Methods

This narrative review comprehensively searched PubMed, PubMed Central, Embase, Web of Science, Scopus, and Google Scholar through May 15, 2026, using all relevant keywords. Peer-reviewed studies reporting AI-based neural or fascial plane recognition applicable to pelvic nerve protection during rectal cancer surgery were included. The manuscript is a narrative review; no protocol was registered, as systematic review methodology was not employed.

Results

The AI Neurorecognition System (AINS) achieved a mean intersection-over-union (mIoU) of 0.75 and 96.09% overall accuracy, reducing nerve recognition time from 25 to 3 min. The Eureka navigation system improved trainee recognition in 20.7–90.6% of missed scenarios without increasing complications. Fascial plane models achieved Dice coefficients of 90.4–90.6%. Robotic micro-space dissection demonstrated acceptable oncologic efficacy with reduced dysfunction. The NEUROS multicenter randomized controlled trial demonstrated that intraoperative neuromonitoring (IONM) significantly reduced urinary and sexual dysfunction compared with standard TME at 12 months — the highest-quality evidence currently available for any nerve-preservation adjunct, providing a validation benchmark for future AI trials.

Conclusion

AI-guided neural recognition has progressed to clinically deployed systems demonstrating high technical accuracy and significant educational benefits. However, no randomized trial has yet demonstrated that AI guidance reduces patient-level functional dysfunction. The transition to standard practice requires multicenter randomized controlled trials measuring validated functional outcomes, head-to-head comparisons with existing nerve-preservation strategies (particularly IONM), rigorous external validation, formal cost-effectiveness analyses, and clear regulatory frameworks.