Scholarlink

Background <p>Intraoperative fluorescence molecular imaging (FMI) is increasingly used to distinguish benign from malignant tumors; however, existing quantitative methods are predominantly data-driven, requiring large datasets and exhibiting limited adaptability in rare diseases or low-incidence tumor surgeries. Furthermore, high false-positive rates and substantial inter-operator variability remain significant challenges, hindering further application of FMI-guided surgery. This study introduces a novel, self-referenced clustering framework that overcomes these limitations by leveraging internal tissue controls and rapid, data-independent analysis.</p> Methods <p>We developed a two-stage clustering framework (<i>K</i>-means followed by Fuzzy C-Means, <i>K</i>-FCM) to process and analyze the tumor FMI results in NIR-II spectrum (wavelength 900–1880&#xa0;nm). Unlike conventional models, our method utilizes each patient’s adipose tissue as an internal reference, enabling individualized thresholding without reliance on large external datasets. This strategy was validated on 16 patients undergoing orbital tumor resection (with injecting indocyanine green, ICG). The diagnostic performance of the proposed framework was compared with the traditional tumor-to-normal ratio (TNR) thresholding strategy.</p> Results <p>The self-referenced <i>K</i>-FCM clustering framework demonstrated substantial improvements over the TNR method, achieving higher sensitivity (0.909 vs 0.818), specificity (0.800 vs 0.600). The use of internal tissue controls effectively normalized fluorescence variability, minimized ICG-related false positives, and enabled accurate intraoperative differentiation of malignant tumors. Furthermore, the proposed framework does not require a large amount of clinical data for training, so it has greater clinical practicality and can be used even for rare tumors.</p> Conclusions <p>Our self-referenced, data-independent clustering framework provides fast and reliable intraoperative analysis for fluorescence-guided tumor navigation. By lowering false-positive rates and the dependency on operator experience, the method enhances the accuracy and promotes the clinical application of FMI-guided tumor surgery.</p> <p><i>Trial registration:</i> On 14th November 2020, the study was registered in the Chinese Clinical Trial Registry (ChiCTR2000039908) and the data shown herein are part of this study.</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

Data-independent fluorescence molecular imaging analysis: a two-stage clustering framework for intraoperative tumor imaging

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月11日

查看全文
  • Yixiang Zhou,
  • Jiaqi Tang,
  • Jie Liu,
  • Yiyin Zhang,
  • Hanfu Shi,
  • Huaping Wu,
  • Wei Wu,
  • Zeyu Zhang

摘要

Background

Intraoperative fluorescence molecular imaging (FMI) is increasingly used to distinguish benign from malignant tumors; however, existing quantitative methods are predominantly data-driven, requiring large datasets and exhibiting limited adaptability in rare diseases or low-incidence tumor surgeries. Furthermore, high false-positive rates and substantial inter-operator variability remain significant challenges, hindering further application of FMI-guided surgery. This study introduces a novel, self-referenced clustering framework that overcomes these limitations by leveraging internal tissue controls and rapid, data-independent analysis.

Methods

We developed a two-stage clustering framework (K-means followed by Fuzzy C-Means, K-FCM) to process and analyze the tumor FMI results in NIR-II spectrum (wavelength 900–1880 nm). Unlike conventional models, our method utilizes each patient’s adipose tissue as an internal reference, enabling individualized thresholding without reliance on large external datasets. This strategy was validated on 16 patients undergoing orbital tumor resection (with injecting indocyanine green, ICG). The diagnostic performance of the proposed framework was compared with the traditional tumor-to-normal ratio (TNR) thresholding strategy.

Results

The self-referenced K-FCM clustering framework demonstrated substantial improvements over the TNR method, achieving higher sensitivity (0.909 vs 0.818), specificity (0.800 vs 0.600). The use of internal tissue controls effectively normalized fluorescence variability, minimized ICG-related false positives, and enabled accurate intraoperative differentiation of malignant tumors. Furthermore, the proposed framework does not require a large amount of clinical data for training, so it has greater clinical practicality and can be used even for rare tumors.

Conclusions

Our self-referenced, data-independent clustering framework provides fast and reliable intraoperative analysis for fluorescence-guided tumor navigation. By lowering false-positive rates and the dependency on operator experience, the method enhances the accuracy and promotes the clinical application of FMI-guided tumor surgery.

Trial registration: On 14th November 2020, the study was registered in the Chinese Clinical Trial Registry (ChiCTR2000039908) and the data shown herein are part of this study.