Encrypted login | home

Program Information

Machine Learning Role in Radiomics and Radiogenomics

no image available
S Gulliford
no image available
G Pandey
no image available
C Fuller

S Gulliford1*, G Pandey2*, C Fuller3*, (1) Institute of Cancer Research & Royal Marsden, Sutton, UK, (2) Mount Sinai, New York, NY, (3) MD Anderson Cancer Center, Houston, TX


4:30 PM : Outcomes models with machine learning - S Gulliford, Presenting Author
5:00 PM : Radiogenomics with machine learning - G Pandey, Presenting Author
5:30 PM : Clinical applications with machine learning - C Fuller, Presenting Author

MO-F-605-0 (Monday, July 31, 2017) 4:30 PM - 6:00 PM Room: 605

Recent years have witnessed exponential growth in patient specific information in radiotherapy ranging from clinical demographics, dosimetrical planning, delivery options, imaging (radiomics), and biological markers (genomics, proteomics, metabolomics) which are contributing to the making of radiation therapy Big data revolution. In particular, there has been increased interest in using dosimetric information in combination with imaging features (radiomics) and/or genetics (radiogenomics) to predict response to radiotherapy. This data revolution provides clinicians with wealth of information on individual patients but also with impeding challenges of making sound reasoning or inferences using traditional statistical methods, which are unfortunately limited by the most frequent patterns and correlative analyses. As result, these methods lack the ability to provide sufficient predictive power when applied prospectively to personalize treatment regimens. In contrast, artificial intelligence techniques based on machine learning (e.g., neural networks, decision trees, random forests, support vector machines, etc.) enjoy the ability to learn from current environment and generalize into unseen tasks. Subsequently, machine learning (ML) methods have been adopted as the modern data analytics vehicle of choice and have recently witnessed unprecedented surge across multiple disciplines from commerce to biomedicine. This is due to their ability to detect complex patterns in heterogeneous datasets with superior results when compared to state-of-the art in each of these disciplines. However, proper application of these methods is not without controversy regarding over-fitting pitfalls and black box stigma. In this symposium, a panel of experts will discuss key insights underlying machine learning methods, the kind of problems they can/cannot solve, how they can be applied effectively to radiation oncology, and what issues are likely to arise in clinical implementations.

Learning Objectives:
1. Perspective on machine learning application for decision support in radiotherapy from clinical, biological and physics.
2. Machine learning role in modern radiogenomics.
3. Machine learning Dos, Don'ts and Maybes.


Contact Email: