PhD Studentship: Machine Learning for the Detection of Clinical Deterioration in Respiratory Patients

Start date: February 2024

Supervisor: Prof Tricia McKeever, Dr Grazziela Figueredo

Patients in hospital can deteriorate for several reasons including sepsis or respiratory failure. Early detection and treatment of patient deterioration can improve clinical outcomes. Current methods of detecting deterioration include intermittent measurement of vital signs such as heart rate and temperature, and the use of simple physiological scoring systems such as the National Early Warning Score-2 (NEWS-2). These methods can miss some cases of deterioration and have a high rate of false alarms, which leads to inefficient use of medical resources, and alarm fatigue among medical and nursing staff. Developing improved methods for detecting patient deterioration is an active field of research. Promising approaches include the application of machine learning techniques to vital signs data, and the use of wearable devices to record continuous physiological data. Since 2015, Nottingham University Hospitals NHS Trust (NUH) has recorded vital signs using an electronic task management system. Approximately 10 million sets of vital signs have been recorded using this system, and this large dataset is available for analysis. More recently, wearable respiratory rate monitors (Respirasense, PMD Solutions, Cork Republic of Ireland) have been introduced on two respiratory wards at NUH, providing a source of continuous physiological data for analysis.

The PhD student will utilise a variety of machine learning techniques including novelty detection, decision tree models and fuzzy logic in order to develop tools for the detection of clinical deterioration, focusing on a population of patients with respiratory conditions. This will be carried out using traditional intermittent vital signs, as well as data recorded using wearable respiratory rate monitoring devices. Machine learning models will be developed in close collaboration with practicing clinicians. Clinical expert knowledge and feedback on existing machine learning models will be gathered via focus groups and questionnaires. This expertise will be modelled using fuzzy logic inference, to be subsequently embedded into the intelligent decision systems to moderate the machine learning outcomes. For clinicians to trust and use the outputs of machine learning models, they need to be able to understand why the outputs have been generated. The PhD student will develop improved techniques for providing human-understandable explanations for the predictions made by the models. These will be improved through feedback from focus groups of practicing clinicians.

Funding notes

The three year studentship covers tuition fees and a tax-free stipend. UK STUDENTS ONLY

Entry requirements:

At least a 2.1 Honours degree in Computer Science, Data Science, Statistics, Mathematics or a related field.

Informal inquiries can be made to tricia.mckeever@nottingham.ac.uk

Application details: to apply for this PhD opportunity, please submit the following documents to tricia.mckeever@nottingham.ac.uk

• Cover letter outlining your research interests and motivation to pursue this project.
• Curriculum vitae detailing your academic background, research experience, and relevant skills.
• Academic transcripts of your previous degrees.

Advert information