The Computer Aided Theragnostics (ukCAT) project aims to develop predictive models using routine clinical data which will, in the future, underpin decision support systems for use in clinics. The use of routine data from all patients, as opposed to selected groups of patients (for example, those that take part in clinical trials) ensures that the results of our analyses apply to everyone, including those patients who rarely take part in clinical trials (e.g. the very elderly and those with other serious medical conditions). We aim to extend ‘big data’ analysis to the complex clinical and imaging data collected every day during routine radiotherapy treatment in the NHS.
The Christie is the first UK partner in the CAT rapid learning oncology network (see www.eurocat.info for details of other international CAT partners) which aims to introduce decision support into clinical practice in the future. The ukCAT project uses a system called ‘distributed learning’ to analyse data.
Information learned from anonymous databases located at hospitals or other medical institutions are securely combined to allow knowledge to be shared in a secure and private environment. Safely pooling information allows models to benefit from using larger numbers of patients (thereby capturing rarer events) and to compare different methods of treatment around the world.
All data held in the ukCAT database is fully anonymised and held on secure servers in accordance with the law, and NHS Information Governance and Research Governance policies. The process of anonymising the data is performed before the data is sent to the ukCAT system meaning it is not possible for a research team member using the system to access any identifiable information.
The central CAT data security policy is that all data on any CAT server is anonymous, and that during distributed learning no data ever leaves a hospital or institution’s own servers - Models go to the data, the data never leaves the institution. The following video developed by our euroCAT partners explains how this works.