Next MICCAI Industrial Talk: AI for Fast Radiotherapy Planning - June 5, 2025
Monday 26th May 2025
Don't miss the next exciting talk in our MICCAI Industrial Talk Series
AI for Fast Radiotherapy Planning
Speaker: Riqiang Gao, PhD, Staff AI Scientist, Siemens Healthineers
Thursday, June 5, 2025
10:30 am (EDT) / 2:30 pm (UTC) / 10:30 pm (CST)
In this MICCAI Industry Talk, Riqiang Gao, PhD, Staff AI Scientist at Siemens Healthineers, will present how advanced AI provides great opportunities to accelerate the process to make radiation therapy available to more cancer patients.
Abstract
Radiation therapy (RT) is a critical treatment modality used in approximately 50% of all cancer cases. However, RT planning is a complex and highly specialized process that demands close collaboration among clinical experts and often involves extended periods of interaction—for example, 3 to 6 hours of active work spread over 1 to 5 days. Advanced AI provides great opportunities to accelerate the process to make RT available to more cancer patients.
First, we introduce our work towards precise 3D dose prediction for radiotherapy (CVPR 2023). Given the CT, planning target volume (PTV) and organ-at-risk (OAR) masks, and other potential plan configurations, fast dose prediction can quickly generate whole 3D dose for planners to review in less than 1 second. In this work, we enable the model to take different planning modes (IMRT vs. VMAT) and variable beam geometries.
Second, we explore a fluence prediction model with deep learning for VMAT planning (Medical Physics, 2025). Fluence map generation traditionally involves time-consuming processes. Our approach enables ultrafast VMAT planning by predicting all the fluence maps of a VMAT arc in one single network.
Third, leaf sequencing—the last mile for the deliverable plan—is predominantly addressed by optimization-based approaches. We propose a novel deep reinforcement learning model in a multi-agent framework for leaf sequencing (ICML, 2024). Our model shows potentially faster convergence when integrated into an optimization planner and promising results in a preliminary full artificial intelligence planning pipeline.
Additionally, to scale up AI modeling with high-quality data, we propose an automatic, scalable solution for generating high-quality treatment plans (arXiv, NeurIPS 2025 submission). Committed to public research, the first data release of our pipeline includes nine cohorts covering head-and-neck and lung cancer sites to support an AAPM 2025 challenge and more RT research. This dataset features more than 10 times the number of plans compared to the largest existing well-curated public dataset to our knowledge.
Speaker's Biography
Riqiang Gao is a Staff AI Scientist in Digital Technology and Innovation (DTI) at Siemens Healthineers. He is a project manager and technical leader in radiotherapy planning research at DTI. He obtained his Ph.D. in computer science from Vanderbilt University. As first author, he has published papers in leading conferences and journals about interdisciplinary fields of machine learning (e.g., ICML), computer vision (e.g., CVPR), medical imaging (e.g., MICCAI, Medical Image Analysis), and clinical validations (e.g., Radiology AI). He has received multiple awards, including the Innovation Excellence Award at Siemens Healthineers, the C.F. Chen Best Student Paper Award at Vanderbilt, and twice RFW All-Conference Best Student Paper Finalist at SPIE medical imaging. His focus on radiotherapy and medical imaging with AI reflects his dedication to advanced patient care.