Speaker : John A. Lee.
January 25, 2021 – 1PM (English)
Radiation Oncology (RO) and Proton Therapy (PT) are highly computerized modalities to treat patients for cancer.
In particular, RO relies heavily on medical imaging to plan an optimal treatment where beams are carefully aimed at the tumor.
To some extent, it amounts to solving a ballistic problem that reaches the best balance between maximum probability of tumor local control and minimum probability of side effects.
Solving this optimisation problem has been partly automated over the years, especially to adjust irradiation parameters like beam angles, shapes, and fluences.
On the other hand, satellite aspects of the problem have remained mostly manual until very recently, when AI started to revolutionise image processing with convolutional neural networks.
Organ delineation on scanner images and treatment dose prediction are the main steps where AI can help radiation oncologists in their workflow.
AI can also intervene in treatment option assessment and selection, with various applications like smart but efficient dose map denoising, image-to-image translation, and clinical decision support systems.
IREC/MIRO and ICTEAM/ELEN are active in these topics and specifically in their application to PT, in collaboration with IBA.
Despite the interest they raise and their contributions in the clinical workflow, these applications still belong to weak artificial intelligence, leaving complex, contextualised decisions to the physician.
Stronger, more integrated, more context-aware AI is yet to come in RO, leaving many research questions open, like medical record processing with mixed data (images, natural language, patient scores and indicators) and interactive reporting after decision-making.