Towards the Use of Probabilities for Treatment Prescription, Optimisation,
and Evaluation of Radiation Treatment Plans
Chairs:
- Edmond Sterpin, Medical Physicist, Katholieke Universiteit Leuven, Leuven (BE)
- Thomas Bortfeld, Medical Physicist, Harvard Medical School, Boston (USA)
Invited speakers:
To be confirmed
Motivation:
Uncertainties are involved at every step of the preparation and execution of a radiotherapy treatment. This concerns, among others, uncertainties in the location of the tumor and its microscopic extension, geometric and anatomical variability on all relevant time scales (breathing, weight loss, morphological changes…), and other physical uncertainties like the ones related to the conversion of images into physical quantities needed for dose computations.
The usual way to handle uncertainties in a quantitative way is to determine probabilities, which are quantified uncertainties. Convenient and widely used formalisms in treatment planning include the notions of GTV, CTV, and PTV. While it may not be immediately apparent, the CTVs and PTVs require probabilities for their determination - either implicitly or explicitly. For instance, PTV margins may be determined by margin recipes, which compute the margin size according to a minimum dose level for a certain patient population coverage under a statistical model.
Those notions are now very familiar and handy. We use them all the time, without even thinking about their capabilities and limitations. However, the limitations can have a profound impact. One example is in proton therapy, where the instability of the dose distributions in the presence of treatment errors has caused the progressive abandonment of the PTV in favor of “robust optimization” on top of the CTV. Robust optimization is also suggested in radiotherapy with X-rays, for instance for breast treatments, because of the difficulty of the PTV concept to handle distortions of the dose distributions.
Another issue, for which solutions are still lacking, is the difficulty to perform consistent trade-offs between PTV coverage and healthy tissue sparing. Most often, those trade-offs are made based on the expertise of the clinicians involved, for lack of anything better, which creates more variability between the clinicians within a center and between centers.
Introducing formalized and explicit probability concepts at every step of the treatment planning workflow could help solve those issues. By defining probabilistic treatment objectives, optimizing the treatment according to them, and evaluating the treatment also probabilistically, we would open an era where trade-offs will always be made in a quantitative fashion. Moreover, new methodologies for integrating probabilities in the optimisation process could be developed, for enabling more conformal treatments. For instance, tumor presence could also be handled probabilistically instead of using fixed CTV volumes like we do today.
Workshop focus:
We will perform a quick overview of the state-of-the art in the following topics:
- definition of probabilistic treatment aims (interpretation of multi-modality imaging, probability maps, target coverage objective, organs-at-risk constraints)
- probabilistic and robust treatment plan optimization
- probabilistic and robust treatment plan evaluation
This overview of the state-of-the-art includes demos of probabilistic planning solutions.
After that, we will identify the key missing elements in the state-of-the-art that prevent a rapid deployment of probabilistic treatment planning, and evaluate which ones could be handled at the level of the envisioned working group.
Potential outcomes:
- Development of a network around probabilistic planning research
- Working group on the integration of probabilistic planning in clinical practice
- Publication about the status of probabilistic planning and vision for the future
- Formalizing the concept of probabilities in treatment planning with scientific rigor, making a clear distinction between marginal and joint probabilities