Room: Track 5
Radiation therapy, using physical tool (i.e., ionizing radiation) to achieve biological effect (i.e., reproductive cell death), is highly quantitative in clinical setting. Outcome modeling and response prediction aims to use of mathematics, physics, biology, and computer science to study the behavior of tissue response to radiation by mathematical modeling and computer simulations. Original radiobiological models, i.e., the linear-quadratic model, TCP/NTCP, EUD, largely based upon mechanic or phenomenological principles on the data, can be used to macroscopically describe radiotherapy-induced dose response. In contrast, analyses of quantitative features extracted from multi-imaging modalities (Radiomics) and their association with molecular features (Radiogenomics) are expected to uncover underlying tumor phenotype for personalized treatment. However, the underlying insights are currently unknown and can’t be explicitly programmed. Empowered by artificial intelligence (AI) and machine learning (ML) methods, radiomics and genomics analyses allow to develop better prediction models to potentially improve diagnostic, prognostic and predictive accuracy of treatment response in RT.
This session will start with introduction of analytical approach in dose-response modeling, and discuss radiomics/genomics analyses using machine learning (ML) and deep leaning (DL) in the context of response assessment to guide personalized treatment. Lastly, the applications and challenges will be discussed.
1. Understand current approaches of outcome assessment in radiation therapy
2. Understand machine learning and deep learning methods in response prediction in RT
3. Understand applications in treatment planning and challenges of treatment outcome modeling in RT