Brendan Delaney

Brendan is a faculty member at I-X, Imperial’s cross-faculty Initiative in applied Artificial Intelligence. He also Chairs the Faculty of Medicines AI Research Ctte. His work covers Artificial Intelligence in medical diagnosis and learning health systems. He is a member of the Medical Research Council Data Science Strategic Advisory Group, and the NIHR MRC Better Methods Better Research oversight group. He is a GP one day a week. He is a leading exponent internationally of the “Learning Health System’ (LHS) concept. Although his initial training in research was in heath technology assessment, real-world (pragmatic) clinical trials and clinical research in Family Medicine, since 2003 he has worked in the area of Clinical Informatics, being appointed to a Chair in Medical Informatics at Imperial in 2015 and elected one of the first 100 founding fellows of the new UK Faculty of Clinical Informatics in 2017 (Now Health and Social Care Faculty of the British Computer Society). He has had wide exposure to European and US clinical informatics through workshops and symposia.

Talk Title

AI in Medical Diagnosis

Talk Summary

My interests lie at the intersection of health services research (how to deal with patient problems equitably and efficiently), data semantics and clinical meaning, and machine learning based model validation and evaluation (AI). Currently there are two areas of active research:

1. Cancer diagnosis in Primary Care. I have led a number of projects in the area of multi-modal data integration, analytics and model validation with a view to generate ‘start of the consultation’ differential diagnosis lists to improve the timeliness and accuracy of cancer early detection in primary care. We have just been awarded a £2.5m grant from the NIHR i4i programme in collaboration with SOAP Health, a US startup with an AI virtual doctor to take an unbiased medical history and provide prompts for diagnosis prior to a clinical encounter.

2. Computable Clinical Guidelines and Explainable AI. Completing the LHS cycle with a computational infrastructure for deploying guidelines as decision support linked to the EHR. Initially supported by an EPSRC Global Health Development Project (www.ROAD2H.org), this area is now supported by Imperial College NHS Trust Biomedical Research Centre. The aim is to both trigger individual guideline recommendations and add explainability.