DeepCardio: towards a better prediction of heart attacks and stroke using a machine learning approach

Researchers: James Rudd, Elizabeth Le, Yuan Huang, John Aston, Emanuele Di Angelantonio (Public Health and Primary Care), Mihaela van der Schaar (University of Oxford, ATI)

The UK Biobank is multidimensional dataset of 500,000 volunteers comprising imaging, genomic and circulating biomarkers. We are applying machine learning techniques to this dataset to predict first heart attack, and will extend this work to include both atrial fibrillation and stroke. Van der Schaar’s group have developed an AutoPrognosis pipeline which enables the data to be fitted to numerous different off-the-shelf machine learning algorithms with automatic tuning of the hyperparameters followed by an output of prediction accuracy and calculations of the precision and recall scores.

This project further aims to apply unsupervised machine learning to the wealth of variables available in the UK Biobank to extract novel insights into which variables may have importance in the determination of disease.

Who's involved