Session Item

We are currently experiencing a revolution in the world of radiation oncology that will drastically change the activities of the medical physicists in the short term. Activities are changing, moving from instrumentation, metrology, dose calculation, simulation activities that were the core of the radiation oncology medical physicist's job until now to much more multidisciplinary research areas. These shifts will change in the short term the medical physicist’s curriculum, requiring the need to acquire new interdisciplinary skills that will allow him/her to position himself/herself on many innovative topics and stay in the game. These skills include, but are not limited to, applied mathematics, radiobiology, or complex system modelling. In this context, programming has already become a prerequisite for the medical physicist profession in recent years.  

Among them, artificial intelligence is becoming indispensable in the radiotherapy workflow, from OAR and Target volumes segmentation to adaptive radiotherapy implementation, the ultimate step being the deployment of precision radiotherapy. Industrials have rapidly positioned themselves on the market and are proposing tools that are already finalized or in the process of being developed for certain stages of the workflow, demonstrating the real impact of these models in the clinic. The first step was automatic contouring, which until then required discussions with physicians. The next steps concern planning on synthetic images, treatment planning or quality assurance, and will require constructive inputs from medical physicists. We are therefore reaching a critical moment where medical physicists must be trained in data sciences in order to continue to be able to discuss with industrials, and to keep a critical mind with regard to the tools that are proposed to them. In fact, how to know the strengths and weaknesses of an AI software deployed by a company if we don't ask the right questions (evaluation strategy of the generalizability of the model, explainability, etc)? How can we continue to be a force of proposal if we no longer speak the same vocabulary as the manufacturers?

The survival of the medical physics profession depends on being properly trained in AI!