Mining the radiotherapy dose: exploring dose-response patterns in radiation therapy
Chairs
Alan McWilliam: Manchester, UK
Giuseppe Palma, Laura Cella: IBB-CNR, Italy
Invited speakers: Oscar Acosta (Rennes University, France) and Marnix G Witte (AVL, Amsterdam, Netherlands)
Summary:
The conformality rush of the last decade revealed the existence of radiobiological phenomena that were either concealed or disregarded in the classical RT treatment strategy. On the one hand, the progressive sparing of healthy tissue permits to focus on toxicity outcomes that would have been neglected in the economy of past RT modalities. On the other hand, the increasing heterogeneity of dose distribution to OARs highlights unprecedented dose-response patterns and, as a result, emphasizes the limit of the traditional dose-volume histogram (DVH)-based toxicity analysis and toxicity modeling philosophy. The high sparing capability of modern techniques at the same time demands for more and more accurate insights of possible avoidance region within a specific OAR for a knowledge based plan optimization.
A new methodology is emerging where the spatial information of the planned dose for every patient is maintained. No assumptions are made regarding anatomical regions, instead the dose in every voxel across many patients is analysed against a given outcome. This process identifies sub-regions of the anatomy that are most strongly correlated against these outcomes and better defines the anatomy that drives these outcomes.
These techniques have been used to identify anatomical regions that drive biochemical recurrence, defined organ subregions correlated with a given toxicity outcome or with mortality.
To fully enable these techniques, we need robust image normalization approaches as well as statistical analysis. In addition, a crucial issue is the possibility to be able to mine dose distributions across multiple institutions. This would ensure a wider heterogeneity of patient populations and treatment techniques, more robust results and opportunities for validation. Combining datasets can be achieved by pooling data in one location or via distributed learning networks.
Day 1
Introduction to topic: Why do we want to link spatial dose to outcomes
Opening presentations:
The pillars of spatial-dose methods (current methodologies):
(2 / 3 presentations)
- spatial normalisation
- statistical approaches
- building predictive models / current results
What do we need to further implement these across multiple centres?
Welcome and brief introductions from group participants
General pitches by participants (opportunity to showcase work in this field where pitches do not fall into a discussion topic below).
Lunch
Each discussion topic will have an opening ~10-minute presentation to set the scene, an opportunity for pitches by participants, followed by a structured discussion.
Discussion topic 1:
Data (clinical and IT perspectives)
Data collection / standardisation / metadata
Data pooling versus / Distributed learning approach
How do we ensure quality control?
Output: Consensus paper
Day 2
Discussion 2:
Methodologies
Patient characteristics, correct pre-processing steps, non-rigid registration – ensure balance and quality control
Statistical analysis, correction for multiple testing
Output: Formation of consortium – methodology paper
Discussion 3:
Clinical translation
How do we ensure results are translatable into clinical practice? Or actionable clinical hypothesis?
Output: Roadmap paper - clinical validation and adoption
Lunch
Final summary
Summary of discussions and formation of working groups for each discussion topic.
Potential active groups
Manchester University, UK: Alan McWilliam, Andrew Green, Eliana Vasquez-Osorio, Marcel van Herk.
IBB-CNR, Italy: Giuseppe Palma, Serena Monti, Laura Cella.
AVL, Amsterdam, Netherlands: Marnix Witte.
Rennes University, France: Oscar Acosta. E Mylona
University of Cambridge: Raj Jena, Leila E. A. Shelley
MSKCC, New York, USA: Maria Thor, Joe Deasy.
University of Western Australia: M. Ebert
UC San Diego: Moiseenko V