Project in Progress:

Papillary thyroid microcarcinoma (PTMC) is a form of thyroid cancer characterized by a tumor size of less than 1 cm. Effectively managing PTMC while minimizing patient discomfort presents a significant clinical challenge. Radiofrequency ablation (RFA) therapy has emerged as a promising minimally invasive alternative to surgery for PTMC treatment. This technique uses thermal energy to precisely destroy cancerous tissue, offering several advantages over traditional surgical methods and radiation therapy, such as shorter recovery times and the feasibility of outpatient treatment. RFA is an image-guided technique involving two key steps: targeting the PTMC tumor by inserting a needle within the tumor boundary and monitoring the therapeutic procedure to assess the extent of tissue ablation. Currently, ultrasound B-mode imaging is employed to guide RFA, enabling surgeons to identify the tumor boundary and the ablated area by assessing grayscale changes in the images. However, the precision of imaging and targeting, as well as the skill required from physicians, remain significant challenges.

Ultrasound B-mode imaging suffers from limited resolution, which can lead to inaccuracies in identifying tumor boundaries and ablated areas. During RFA ablation treatment, each RF sonication generates microbubbles at the targeted area due to the ablation effect. These microbubbles can persist for minutes to hours, causing acoustic shadowing at the tumor boundary. Consequently, subsequent sonications may face difficulties in accurately identifying the tumor boundary and the treated area. The lack of precise guidance in current ultrasound-guided RFA devices necessitates extensive training for physicians, with therapeutic results heavily dependent on their expertise and experience.

To address these issues, I initiated a research project at the Waterloo AI Institute in collaboration with the Korean Society of Thyroid Radiology. The goal of this project is to develop an AI model integrated with ultrasound imaging for two primary tasks: real-time tracking of the tumor margin and real-time identification of the ablated tumor area during each treatment session. Our team has made significant progress on this project, including the development of novel AI segmentation models integrated with ultrasound imaging for real-time labeling and tracking of tumor margins during treatment and identifying treated areas post-RFA.