PhD Thesis Defense: Sophie Lloyd

ZOOM LINK

"Development of an intraoperative oral cancer detection device and imaging system"

Abstract

The majority of oral squamous cell carcinoma (OSCC) cases are diagnosed in advanced stages, resulting in worse survival outcomes and more invasive surgical treatments that diminish patient quality of life. In resections of these advanced OSCC tumors, it is challenging to successfully remove all cancerous tissue due to infiltrative growth into complex anatomy. Currently, clinicians lack noninvasive early diagnosis tools, and surgeons lack effective methods to intraoperatively assess tumor margins.This dissertation presents the development and clinical evaluation of a novel handheld device leveraging Electrical Impedance Spectroscopy (EIS) and Electrical Impedance Tomography (EIT) for non-invasive, in vivo detection of OSCC and intraoperative margin visualization.

A custom oral impedance probe was designed, fabricated, and characterized for intraoperative use, delivering 0.1 mA current injection across 100 Hz to 100 kHz. An automated data processing pipeline was developed to remove measurements impacted by in vivo artifacts and to aggregate repeated acquisitions to improve signal-to-noise ratio. The full clinical study enrolled 114 patients, yielding 275 validated in vivo samples from 100 patients across four pathology categories: healthy, OSCC, high-grade dysplasia, and benign, constituting the largest reported database of in vivo oral tissue impedance measurements.

Machine-learning and dimensionality reduction methods identified the most informative impedance features using rigorous cross-validation, reducing the feature space to less than 1% of the original dataset while preserving classification performance. Full clinical results classification using impedance magnitude achieved an AUC of 0.92 (84% accuracy) for healthy vs OSCC discrimination and multiclass classification achieved average AUCs up to 0.82. The strong performance of single-frequency parameters offers a translational pathway toward real-time tissue classification and simplified instrumentation. Finally, three novel fused EIT reconstruction methods were developed to spatially visualize OSCC across resection surfaces by combining multiple probe measurements, with fused difference reconstructions demonstrating superior robustness to noise in simulation and phantom experiments.

This work advances impedance-based technologies for oral cancer detection by demonstrating high-accuracy tissue classification under clinical conditions and establishing a foundation for real-time surgical margin visualization, representing meaningful progress toward improving outcomes for patients with OSCC.

Thesis Committee

• Ryan Halter (chair)
• Ethan Murphy
• Joseph Paydarfar
• Michael Kokko
• Emily Porter (McGill) 

Contact

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