PAM was a three year project involving the Universities of Nottingham, Southampton and Stirling and arising from an EPSRC IDEAS Factory on Mobile Health Care. Its aim was to investigate the use of technological aids, based upon wearable and home based sensors together with mobile phones, to help people who suffer from mental health problems. The rationale was that it is well accepted that the early detection of deleterious changes in such people may enable the avoidance of increasingly worsening situations that often lead to the need for admission to hospital. The investigators across the project provided a mixture of the necessary skills and experience across: medical devices and sensors, medical signal processing, communications and software services, and Operational Research modelling.
A steering group consisting of healthcare professionals, including psychiatrists and patients confirmed the opinion that Bipolar Disorder was an appropriate mental health condition to consider. Factors behind this included: its cyclic nature; the usual existence of two clearly differentiated states (i.e. manic and depressive episodes); the self awareness of many sufferers; and severity of impact upon their lives. Via analysis of the research and clinical literature and discussion with the steering group a number of prodromes and syndromes that characterise the illness were identified and then 'mapped' onto sensors that in theory would enable their occurrence to be identified.
This information was used to design, construct and test (on healthy adult volunteers) a monitoring system comprising: a wearable sensor unit, mobile phone, GPS unit (if not available from the phone), 'environmental' sensing unit placed in the subject's home, bed occupancy sensor and door switches and cameras used to monitor activity within the home environment. The wearable and 'environmental' sensing units contained light and sound sensors to monitor ambient conditions.
Technical trials conducted under appropriate ethical approval permitted assessment of the performance of the individual sensors and enabled the development of information extraction algorithms. Analysis of the collected data showed that it is possible to extract activity signatures in an unobtrusive means from everyday activities recorded using such sensors. It is possible to observe repeatable 'activity signatures' from the sensors and so by setting appropriate thresholds it may be possible to detect use detected changes in them to issue alerts.
A limited trial on a volunteer with Bipolar Disorder, under the more detailed appropriate ethical approval, provided a further evaluation in a "Real world" setting. Recruitment of volunteers proved more difficult than anticipated that highlighted the fact that lack of concordance with such a monitoring process may prove an obstacle for widespread adoption; something that a number of psychologists commented upon when the work was presented at conferences. This led to experimentation of a 'lighter' version of the ambulatory monitoring process with more focus upon the mobile phone.
The degree of personalisation was also an important factor in the trials, and so the project has explored issues of programming such systems using rules at run-time. While effective, such a dynamic situation may not result in coherent distributed behaviour. So algorithms have now been developed to both detect and resolve such issues at run-time; however this aspect of personalisation has still to be exercised within a live trial.
During the project a number of links to potential future collaborators were established including academics within the EU, the University of Nottingham Institute for Mental Health and several commercial entities. The project team is actively looking to advance this work further by collaborating with existing PAM partners and these new contacts with appropriate funding schemes and further papers.