Bookmarks
ID-529: Deep Learning Models for Cancer Risk Prediction
Safe People
Organisation name
Imperial College London
Applicant name(s)
Brendan Delaney
Funders/ Sponsors
Safe Projects
Project ID
ID-529
Lay summary
Develop and validate transformer AI using WSIC EHR data to predict cancers (pancreatic, lung, colorectal). Models aim to identify at-risk populations, inform screening, and improve understanding of disease progression and risk factors
Public benefit statement
Professional benefits include supporting clinicians, particularly GPs and pharmacists, in making more informed and individualised decisions on cancer referrals. It could improve confidence in strengthening shared decision-making with patients, and contribute to more consistent implementation of NICE guidance. The project may also help build analytic capacity among healthcare professionals working with linked real-world data, prediction models, and medicines safety evaluation. Academic benefits are substantial. The project would advance methods for diagnostic foundation models, causal inference, model explainability, linked EHR analysis, and adverse event prediction using real-world data. The framework developed could be adapted for creating opportunities for further grant applications, methodological publications, cross-disciplinary collaboration, and future research in cancer diagnosis, prevention, and clinical AI. Commercial benefits may arise through the development of transferable risk prediction tools, analytic pipelines, or decision-support approaches that could be implemented in health technology products. The methods and software could be of interest to life sciences companies, health data platforms, digital health companies, or NHS innovation partners interested in post-marketing surveillance, stratified prevention, and medicines optimisation. The project could also help de-risk later investment in scalable clinical decision support tools by demonstrating feasibility and value in a real-world NHS setting. Organisational benefits include improving how healthcare systems monitor and manage cancer risk across care settings. By linking primary and secondary care data and reconstructing patient pathways, the project could strengthen integrated care approaches, improve identification of high-risk patients, and support earlier intervention. For NHS organisations and integrated care systems, this could inform service planning, prevention strategies, and resource allocation. It may also provide a reusable framework for evaluating other drug safety questions, supporting governance, quality improvement, and evidence generation for commissioning and reimbursement decisions.
Other approval committees
Project start date
18/06/2026
Project end date
21/12/2026
Latest approval date
18/06/2026
Safe Data
Dataset(s) name
Data sensitivity level
De-Personalised
Release/Access date
25/05/2026
Safe Setting
Access type
TRE