PhD Thesis Defense: Yao Chen

Optional ZOOM LINK
Meeting ID: 954 8504 4452
Passcode: 204092

"Advancing Computational Analysis of Fluorescence Imaging for Head and Neck Cancer Surgical Guidance"

Abstract

The primary objective of surgery for head and neck squamous cell carcinoma (HNSCC) is complete tumor resection while preserving critical normal structures to optimize survival and functional outcomes. Achieving negative margins remains challenging due to complex anatomy, infiltrative tumor growth, and the limitations of visual identification. Near-infrared (NIR) fluorescence molecular imaging targeting epidermal growth factor receptor (EGFR) has emerged as a promising adjunct to surgery and has driven the development of fluorescence-guided surgery (FGS). While early clinical trials demonstrate feasibility and safety, modest tumor-to-normal contrast and variability in quantitative interpretation have limited broader clinical adoption.

These limitations arise from both biological and computational challenges. Physiologically, heterogeneous tumor delivery, limited tissue penetration, receptor binding dynamics, and systemic pharmacokinetics influence the spatial distribution of imaging agents and the resulting imaging contrast. Computationally, robust frameworks for fluorescence-based tumor interpretation remain underdeveloped, and contrast assessment methods often lack standardization across institutions and imaging systems. Addressing these challenges requires a comprehensive framework that integrates data-driven analysis with physiology-informed modeling to improve the interpretability and reliability of fluorescence imaging in surgical decision-making.

This thesis work advances computational analysis of fluorescence imaging through three primary components. First, machine learning-based "fluorescence optomic" analysis was developed to extract quantitative image features and improve tumor detection and margin assessment accuracy in clinical HNSCC datasets. Second, a multi-institution contrast-analysis framework was established to evaluate cross-site reproducibility, standardize region-of-interest selection and contrast computation, and systematically compare contrast metrics for generalizable margin distance estimation. Third, a physiology-grounded computational modeling framework was constructed to simulate microscale agent transport and tissue penetration dynamics, enabling mechanistic evaluation of agent parameters and dosing strategies on tissue distribution and resulting imaging contrast. Collectively, these advances provide an integrated pathway toward more quantitative, standardized, and translational fluorescence-guided surgery.

Thesis Committee

• Kimberley S. Samkoe (Chair)
• Scott C. Davis
• Geoffrey P. Luke
• Greg M. Thurber (U Michigan)

Contact

For more information, contact Thayer Registrar at thayer.registrar@dartmouth.edu.