PhD Doctor of Philosophy
Improving patient mortality through enhanced prediction with vital signs data
We will use Artificial Intelligence techniques and in particular deep learning to predict deterioration of patients in Healthcare using data that is already routinely captured in the UK as part of the (National) Early Warning Score (NEWS). The aim is to provide significantly improved and accurate prediction of health deterioration compared with the existing NEWS procedures. This would allow earlier intervention with the potential of significantly improving clinical outcomes. The poor quality of the data available will present a particular challenge as will the development of all the processes required to create an end to end pipeline. We will also determine if other data, other than currently used, would further improve prediction capabilities. A significant effort for the Thesis will involve benchmarking of different types of algorithms including novel implementations, combinations of algorithms, and finding new ways for improvements and tuning of algorithms. Coupled with this will be the development of processes to visualize multi-modal time-series data, identify problems in data quality and create processes to compensate for irregularities including missing data, data taken at different time intervals and incorrect entries.