Skip to main content

Postgraduate research project

Kernel methods for system identification with application to the analysis and classification of cardiovascular time series

Fully funded (UK only)
Type of degree
Doctor of Philosophy
Entry requirements
2:1 honours degree View full entry requirements
Faculty graduate school
Faculty of Engineering and Physical Sciences
Closing date

About the project

An emerging research trend in cardiovascular sciences focuses on discovering useful diagnostic information from the dynamic analysis of concurrent physiological measurements, such as the electrocardiogram (ECG), continuous blood pressure, cerebral blood velocity, CO2 levels (capnography) etc.

The aim of this PhD project is to develop new methods for the analysis of time series cardiovascular measurements. This will lead to faster and better targeted treatments for conditions such as stroke and head injury, resulting in improved brain protection and better outcomes for patients.

System identification denotes the task of building mathematical models of dynamical systems starting from time series of input and output data. It can be seen as a generalisation of learning a functional relation but where the output depends also on past inputs. It includes building “black-box” models, but also the validation of physics-based models. System identification is often rooted in the characterisation of dynamical systems studied in control engineering. This gives system identification a distinctive mathematical background and an advantage in applicative domains including aerospace, processes, and biomedical engineering.

System identification has been mostly approached using statistical parametric estimation. However, in recent years, a new machine learning approach to system identification has emerged based on kernel and regularisation techniques. This approach offers advantages over classical statistical methods, especially when using real-world data where less strict assumptions give more degrees of freedom for tuning the solution. This PhD project aims to contribute to the development of this approach and take it in new directions and into new applications in the biomedical field. Developments of particular interest are the extensions to multivariate systems and to population formulations where data from similar but different systems are combined, the latter being particularly relevant to biomedical data.

The project will have a focus on blood flow and its physiological control in the brain. In this area, the project will benefit from the availability of physics-based models, and of real-world recordings from healthy subjects and clinical patients and includes collaboration with clinical partners at Southampton General Hospital. The project is also linked with the activities of the international Cerebrovascular Research Network CARNet.

to top