BAPTON - Biology Adaptive for Proton Therapy in Oesophageal caNcer

More than half of the patients treated for cancer receive external Beam Radiation Therapy (EBRT) during the course of their disease. EBRT is typically characterized by highly standardized protocols and relies on the precise daily delivery of high energy X-rays combined with volumetric imaging tools. In order to further improve cancer care in Belgium, the UZ Leuven and the Cliniques Universitaires Saint-Luc have recently built in partnership the first Belgian proton therapy center, called ParTICLe, which is located on the Gasthuisberg campus in Leuven. The chosen manufacturer was IBA, a former spin-off from UCLouvain in 1986 and now the world leader in the construction of proton therapy centers. The first patient is expected by March 2020. Proton therapy (PT) sets itself as a major player in the therapeutic arsenal available to clinicians on both sides of the linguistic border to treat their patients with one of the most advanced technologies. In comparison to photon therapy with X-rays (XT), proton physics allows a high concentration of the dose delivered in the target volumes resulting in a better sparing of the surrounding healthy tissues and a better therapeutic ratio. However, the added value of PT is not systematic and depends on anatomical and clinical parameters which are specific to each patient. In addition, patients with similar baseline profiles may respond differently to treatment. So, a patient under treatment with XT might develop unexpected toxicities which urges switch to PT. Optimal usage of such advanced technology therefore requires to determine, before and during treatment, the patients who could benefit the most from PT. The aim of the project is to develop a clinical decision support system, largely automated by artificial intelligence, which allows to objectively assess the added value of PT compared to XT at any time during treatment, for every individual patient. This assessment will be accomplished using published toxicity and tumor control prediction models that will be updated based on longitudinal imaging data and biological markers (i.e., acquired during treatment). The tumor site chosen for the proof-of-concept is oesophageal cancer, which is already the focus of several ParTICLe researchers. Furthermore, we will participate in a European randomized clinical trial in which patients with oesophageal cancer will be randomized between PT and XT. The role of artificial intelligence is to address some bottleneck steps in the usual treatment planning procedure, which normally requires human intervention. These manual operations are typically slow, tedious, and prevent fast simulation of the treatment planning chain. Such manual steps are for instance the delineation of organs at risk, the contouring of target volumes, the fine adjustment of the plan objectives to get a clinically acceptable dose.