Enhancing Label-Free Fluorescence Lifetime Imaging for Intraoperative Tumor Margin Delineation in Head and Neck...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate intraoperative tumor margin delineation is crucial for head and neck cancer (HNC) treatment success.
• Label-free Fluorescence Lifetime Imaging (FLIm) offers real-time, contrast-agent-free tissue differentiation.
• Existing FLIm classification models require enhancement for robustness and accuracy in clinical settings.

Purpose of the Study:

• To develop and validate a data-centric AI framework to improve FLIm-based classification of HNC.
• To assess the interpretability of FLIm-derived features for HNC margin assessment.
• To evaluate the impact of patient-specific factors on classification performance.

Main Methods:

• Collected in vivo multispectral FLIm data (355 nm excitation) from 92 HNC patients.
• Implemented a data-centric AI approach using confident learning to refine model training.
• Integrated an interpretability framework to analyze feature contributions and FLIm contrast sources.

Main Results:

• The AI-enhanced FLIm model achieved an Area Under the Curve (AUC) of 0.94 for healthy vs. cancerous tissue differentiation.
• Borderline predictions in tumor regions correlated with transitional tissue properties and metabolic cofactors (NADH, FAD).
• Classification performance was analyzed concerning tumor site and HPV (p16+) status.

Conclusions:

• Label-free FLIm, powered by AI, shows significant potential for real-time intraoperative margin assessment in HNC.
• The technology can guide surgical precision, potentially improving resection completeness and reducing recurrence.
• FLIm-derived metabolic signatures offer insights into tumor microenvironment and aid in dysplasia grading.