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

Area of Science:

• Medical Imaging
• Oncology
• Biotechnology

Background:

• Intraoperative fluorescence molecular imaging (FMI) faces challenges with data-driven quantitative methods, large dataset requirements, and operator variability.
• Existing FMI approaches struggle with adaptability in rare diseases and high false-positive rates, limiting clinical application.
• A novel, self-referenced clustering framework is introduced to overcome these limitations using internal tissue controls and data-independent analysis.

Purpose of the Study:

• To develop and validate a novel, self-referenced clustering framework for intraoperative fluorescence molecular imaging (FMI).
• To overcome limitations of existing data-driven FMI methods, including adaptability and high false-positive rates.
• To enable accurate, individualized tumor differentiation in FMI-guided surgery without large datasets.

Main Methods:

• Developed a two-stage clustering framework (K-means followed by Fuzzy C-Means, K-FCM) for analyzing near-infrared-II (NIR-II) spectrum FMI data.
• Utilized each patient's adipose tissue as an internal reference for individualized thresholding, eliminating reliance on large external datasets.
• Validated the framework on 16 patients undergoing orbital tumor resection using indocyanine green (ICG) and compared it to the traditional tumor-to-normal ratio (TNR) method.

Main Results:

• The self-referenced K-FCM framework achieved higher sensitivity (0.909 vs 0.818) and specificity (0.800 vs 0.600) compared to the TNR method.
• Internal tissue controls normalized fluorescence variability and minimized indocyanine green (ICG)-related false positives, enabling accurate differentiation.
• The framework demonstrated clinical practicality, requiring no large training datasets and showing potential for rare tumor applications.

Conclusions:

• The self-referenced, data-independent clustering framework offers fast and reliable intraoperative analysis for fluorescence-guided tumor navigation.
• This method enhances accuracy in FMI-guided tumor surgery by lowering false-positive rates and reducing dependency on operator experience.
• The framework promotes broader clinical application of FMI in distinguishing benign from malignant tumors.