Time series analysis of medical data

The human body relies on a number of control mechanisms for regulating the flow of blood. Biomedical signals have provided much insight into the different processes that are active in maintaining this regulation. While linear techniques are useful for separating components that operate at different frequencies, nonlinear techniques are required to uncover the complicated interactions that are active within the cardiovascular and cerebral systems. In many cases, pathological disease is expressed as a decrease in the complexity of the fluctuations in the observed signals so that a nonlinear analysis of these signals is capable of facilitating a medical diagnosis.

My aim is to combine knowledge from physiology, usually expressed through nonlinear models, with nonlinear time series analysis techniques. By fitting a physiological model to data recorded from patients it is possible to visualise and diagnose their state of health in real-time. The model prescribes all the nonlinear correlations which are known a priori and acts as a `looking glass' with which to view the dynamics of the patient. Monitoring the time evolution of these parameters will provide a means of assessing whether or not the patient is improving. The development of such a condition-monitoring device could permit early clinical intervention and prevention of pathological disorders. This research will provide

• new metrics for assessing heart rate variability,
• physiologically relevant models of the cardiovascular and cerebral auto-regulatory systems,
• a technique for identifying arrhythmias.

People working in this area within OCIAM are

• Dr Patrick McSharry