Advancing Reirradiation and Treatment Outcome Understanding with Robust Dose Mapping and Causal Inference

Radiotherapy (RT) is a vital treatment for around half of all cancer patients. RT delivers targeted radiation to the patients’ tumour, and incidentally to surrounding healthy organs, which can cause side-effects. To reduce the risk of these side-effects, treatments are optimised for each patient based on their detailed medical images to maximise the dose to the tumour whilst minimising the dose to surrounding healthy organs, creating intricate spatial ‘dose distributions’.
Significant improvements in treatment techniques have resulted on improved patient survival following treatment. However, this also means more patients are experiencing tumour recurrence or new tumours near the original site. The future for these patients can be bleak, with limited treatment option: surgery is often not possible due to frailty, and chemotherapy typically ineffective. Receiving a new course of RT, also known as reirradiation (reRT), is emerging as a promising treatment. However, a major challenge hinders the wider use of reRT: we have a limited understanding of how healthy organs react to multiple rounds of radiation. For example, it is not clear how much total radiation dose organs can tolerate before they cease to function properly, especially as some organs can partially heal from previous radiation. These gaps in knowledge, together with severe side effects seen with older techniques, make doctors cautious about recommending reRT.
To enable safe reRT for more patients, it is essential to estimate the total dose that the patient’s anatomy receives. This requires projecting the radiation from previous treatment(s) onto the current patient anatomy, a process called dose mapping.  Dose mapping relies on aligning corresponding anatomical features visible in the medical images using advanced image-matching techniques known as “deformable image registration algorithms”.  However, patient anatomy can change drastically between treatments, due to factors like aging, effects of previous radiotherapy and/or surgeries. Current methods cannot effectively manage these complex changes and they do not consider the different levels of accuracy required to map intricate dose distributions. These shortcomings result in inconsistent dose estimates, making it difficult to understand the total doses organs can handle and how dose affects organ function. Finally, current methods use statistical prediction models that overlook important factors like the location and timing of radiation treatment. These models are not designed to understand cause-and-effect relationships (unlike causal analysis), so their usefulness for decision making and personalising treatment plans is limited.
This proposal aims to create advanced methods for reliable dose mapping and to introduce (spatiotemporal) causal analysis techniques. This will help to build knowledge on radiation limits to guide future reRT. The aim will be accomplished by achieving the following objectives:

Develop dose mapping algorithms capable of handling significant anatomical changes.
Create dose mapping algorithms that incorporate intricate dose characteristics.
Adapt and implement spatiotemporal causal analysis techniques for reRT.
Generate evidence on acceptable cumulative dose levels below which the risk of developing serious side-effects in the reRT setting is clinically acceptable.

The tools developed in this project will ultimately improve safety of reirradiation, enhancing long-term outcomes for many cancer patients by accurately mapping previous doses onto current anatomy and considering the spatial and temporal aspects of reRT within causal analysis. These innovations will reduce the risk of severe side effects and support more effective treatment planning. Furthermore, the generation of high-quality evidence will guide clinical practice, benefiting patients worldwide.