BIMSA >
Deep Learning at the Frontiers of Science
Enabling Robust, Efficient, and Adaptive Proton Therapy Workflows Using Deep Learning Techniques
Enabling Robust, Efficient, and Adaptive Proton Therapy Workflows Using Deep Learning Techniques
Organizers
Speaker
Yuzhen Ding
Time
Thursday, July 10, 2025 10:00 AM - 11:00 AM
Venue
A3-1-301
Online
Zoom 559 700 6085
(BIMSA)
Abstract
Proton therapy offers superior tumor targeting precision, especially for complex anatomical sites like head and neck cancers. However, its effectiveness is compromised by anatomical variations between fractions, potentially leading to dose delivery errors. Online adaptive radiation therapy (oART) mitigates this by adjusting treatment plans immediately before each session. A critical challenge in oART implementation is the requirement for both fast and accurate dose calculations, typically performed using Monte Carlo methods. We present a deep learning-based denoising framework that enables high-fidelity dose estimation from accelerated, noise-prone Monte Carlo simulations. Patient positioning presents another major challenge. Current image-guidance systems face fundamental limitations: CT-on-rails requires patient repositioning, introducing potential alignment errors, while CBCT delivers additional radiation dose with compromised image quality. We propose a novel deep learning approach to enhance setup accuracy while adhering to the ALARA principle for imaging dose. Together, these AI-driven solutions aim to establish a more robust, efficient, and clinically practical workflow for adaptive proton therapy.
Speaker Intro
Yuzhen Ding received her B.E. in Information Security and M.S. in Electronics and Telecommunications Engineering from Xidian University, China, and her Ph.D. in Computer Engineering from Arizona State University in 2022. She is currently a Senior Research Fellow in the Department of Radiation Oncology at Mayo Clinic, Phoenix, Arizona. Her research interests include computer vision, optimization, and proton therapy, with a recent focus on developing deep learning–enabled online adaptive proton therapy workflows.