Scholarlink

Background <p>Accurate intraoperative tumor delineation is essential for achieving maximal safe resection in glioma surgery. Manual segmentation of intraoperative imaging remains the reference standard but is time-consuming and subject to inter-observer variability. Artificial intelligence–based segmentation has emerged as a potential solution to improve accuracy and efficiency during surgery, yet its comparative performance in the intraoperative setting remains unclear. This systematic review aimed to qualitatively synthesize evidence comparing artificial intelligence–based imaging segmentation with manual segmentation methods for intraoperative guidance during glioma resection.</p> Methods <p>A systematic literature search was conducted in PubMed, Cochrane Library, and ScienceDirect following PRISMA guidelines. Studies were eligible if they involved patients with low- or high-grade gliomas undergoing surgery with intraoperative imaging and directly compared artificial intelligence–based segmentation with manual delineation. Outcomes included segmentation accuracy, segmentation time, extent of resection, neurological complications, and survival. Due to methodological heterogeneity, a qualitative narrative synthesis was performed.</p> Results <p>Five studies met the inclusion criteria. Artificial intelligence–based segmentation demonstrated moderate to high agreement with manual segmentation, with reported Dice similarity coefficients ranging approximately from 0.62 to 0.93. Across studies, AI-driven approaches consistently reduced segmentation time compared with manual methods. Clinical outcomes such as extent of resection, neurological complications, and survival were inconsistently reported and could not be quantitatively synthesized.</p> Conclusions <p>Artificial intelligence–based segmentation demonstrates accuracy comparable to expert manual delineation while improving intraoperative efficiency; however, variability across imaging modalities and tumor characteristics limits generalizability. Further prospective studies are needed to establish its impact on clinical outcomes and routine surgical practice.</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–based segmentation versus manual delineation for intraoperative guidance in glioma resection: a systematic review

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月21日

查看全文
  • Jackson Hakim,
  • Michael Lumintang Loe,
  • Iskandar Japardi

摘要

Background

Accurate intraoperative tumor delineation is essential for achieving maximal safe resection in glioma surgery. Manual segmentation of intraoperative imaging remains the reference standard but is time-consuming and subject to inter-observer variability. Artificial intelligence–based segmentation has emerged as a potential solution to improve accuracy and efficiency during surgery, yet its comparative performance in the intraoperative setting remains unclear. This systematic review aimed to qualitatively synthesize evidence comparing artificial intelligence–based imaging segmentation with manual segmentation methods for intraoperative guidance during glioma resection.

Methods

A systematic literature search was conducted in PubMed, Cochrane Library, and ScienceDirect following PRISMA guidelines. Studies were eligible if they involved patients with low- or high-grade gliomas undergoing surgery with intraoperative imaging and directly compared artificial intelligence–based segmentation with manual delineation. Outcomes included segmentation accuracy, segmentation time, extent of resection, neurological complications, and survival. Due to methodological heterogeneity, a qualitative narrative synthesis was performed.

Results

Five studies met the inclusion criteria. Artificial intelligence–based segmentation demonstrated moderate to high agreement with manual segmentation, with reported Dice similarity coefficients ranging approximately from 0.62 to 0.93. Across studies, AI-driven approaches consistently reduced segmentation time compared with manual methods. Clinical outcomes such as extent of resection, neurological complications, and survival were inconsistently reported and could not be quantitatively synthesized.

Conclusions

Artificial intelligence–based segmentation demonstrates accuracy comparable to expert manual delineation while improving intraoperative efficiency; however, variability across imaging modalities and tumor characteristics limits generalizability. Further prospective studies are needed to establish its impact on clinical outcomes and routine surgical practice.