Hamlet.rt: multi-centre expansion

Background Heuristics, Algorithms and Machine Learning: Evaluation & Testing in Radiation Therapy (Hamlet.rt) is a technology-enabled prospective clinical study aiming to build predictive radiomic models for individualisation of radiotherapy in Head & Neck, Prostate, Brain and Lung cancers. It was established as a single centre study in Cambridge in October 2019.

The study concept arose from a patient focus group, and aims to answer a key question: "Can machine learning models that incorporate large, longitudinal imaging, clinical and patient-reported data, predict radiotherapy toxicity much better than current simple clinical review of individual patient’s radiotherapy plans?" This application relates to a multi-centre expansion of the original prospective clinical study.

Aims This application aims to extend the original study to 10 additional radiotherapy centres, in order to increase the amount of available patient data to over 2,000 patients, enhance the diversity of patients included in the study, and to establish a robust method for cross-validation of predictive models.

In addition, the expansion of the study supports the development of complementary translational sub-studies to evaluate tumour response to radiotherapy.

Methods The study will utilise image analysis, feature extraction and machine-learning based radiomic analysis of radiotherapy image datasets to establish image features that correlate with treatment outcome.

Outcome data is captured through the use of serial electronic patient reported outcome measures completed by patients during radiotherapy and every 3 months for up to 2-years post completion of radiotherapy.

For brain tumour patients, diffusion tensor magnetic resonance imaging (DT-MRI) will also be captured in order to build image based models of white matter connectivity, in order perform spatial correlation of radiation effects with assessment of neurocognitive function. How the results of this research will be used The results of the project will be used in two ways.

The first is to deploy machine learning models that can assist clinicians in identifying patients at increased risk of treatment toxicity from radiation therapy into the clinic, assisting clinicians to individualise treatment for patients, or engage in early interventions to monitor and treat treatment toxicity.

The second is to provide biosamples and high quality clinical outcome data for a series of subsequent translational research studies, that will correlate tumour biology with therapy response.