A combined modelling and machine learning approach to simulate vital signs in critically-ill patients.

The relevant EPSRC accepted research areas are:

Mathematical Biology
Nonlinear Systems
Artificial Intelligence Technologies.

This project will be part of the EPSRC-funded CHIMERA mathematical science in healthcare hub. CHIMERA aims to develop new mathematical models and data science tools which utilise the wealth of physiological data collected for each patient to provide clinicians with a better idea of how well the patient's body is recovering since their admission to ICU. This project will explore various contender models that simulate the mechanical responses and interactions between respiratory and cardiovascular systems and use a combination of analytical, numerical and advanced machine learning techniques to optimise the model structure and fit the parameters to the available clinical data. For the best performing models, large cohorts of virtual subjects will be produced by varying the values of clinically relevant parameters with the purpose of classifying patients into different risk categories of a certain event. Deep-learning architectures will be implemented for each candidate model, given that the classification decision boundaries from model parameters are likely to be non-linear and not simply connected. Our clinical partners will be explicitly consulted to see how/which of the risk classifications obtained might best assist clinical decision making.