ECE MS Thesis Defense: Yuan Sui

Title: Developing a Scanning System for Intraoperative Deep-Margin Assessment of Oral Cancer Tumors Using Elastic Scattering Spectroscopy

Presenter: Yuan Sui

Advisor: Professor Irving Bigio

Committee: Professor Ji-Xin Cheng, Professor Darren Roblyer

Abstract: Oral cancer has a global incidence that varies from one to 22.2 cases per 100,000 people with a high mortality rate of approximately 47%. A typical treatment is a combination of radiation, chemotherapy and surgical intervention. Delayed diagnosis and a high recurrence rate require that complete surgical removal of the tumor includes a margin of healthy tissue surrounding the cancerous tissue in order to reduce the recurrence rate. Frozen sections (FS) are currently used to analyze the margins of the resected tissue intraoperatively, but this technique is time-consuming, requires manual analysis, and has low sensitivity. Elastic scattering spectroscopy (ESS) is a backscattering optical technique based on fiber optics, which can detect subcellular changes in tissue when using a small source-detector separation. When ESS is combined with a tissue-surface scanning function, it will be one solution to the problem of delayed detection of oral cancer during surgical procedures. The current (fourth-generation) ESS system, coupled with a tissue-scanning system, has the characteristics of low cost, miniaturization, and modularity, and can generate 100-micron resolution maps of the returned tissue diagnosis. In the current generation there are drawbacks in clinical use: the main module of the system can only return data and generate data graphs, the scanning function runs for too long, the data processing method is slow, and the system cannot classify healthy and cancerous tissue. The goal of this thesis is to revise the design of the fourth-generation ESS equipment in order to, improve the speed of the scanning module, extend the scanning area, provides a high-resolution image of the scanned tissue, and develop methods for both faster data processing and a machine-learning algorithm to classify healthy and cancerous tissue. A new scanning module has been developed to allow for scanning over an area of up to 25 cm2 with a scan-step as small as 100 um, and it will also provide a photo image for co-registration of the scanned tissue. A diagnostic algorithm will classify the returned data and draw the boundary between healthy and cancerous tissue on the scanned tissue image.