Telemonitoring for Neurological Disease Management
One consequence of an ageing population in the UK is that the incidence of crippling neurological disorders such as Alzheimer's and Parkinson's disease is set to increase, leading to increasing demand on health services [1]. The ability to automatically assess disorders using biological signals is of value in a wide variety of clinical contexts, from pre-screening to the assessment of medical interventions. Of considerable promise is the opportunity to augment visits to the clinician with remote, internet screening and monitoring, particularly for such immobile and often physically disabled patients. Coupled with the substantial lowered cost of computing resources, automated, remote, objective assessment of symptoms present in biological signals, a critical part of the emerging technology known as telemedicine, has recently become a focus of industrial and academic research [2].
Nonlinear and Stochastic Fractal Signal Analysis
In collabortion with Intel Corporation and the National Center for Voice and Speech, obtaining speech signals from patients with Parkinson's disease, this project follows on from an innovative pilot study has shown that, due to the inherent nonlinearity of speech signals [3], mathematical tools such as recurrence analysis (nonlinear repetitiveness) [5] and fractal scaling analysis (statistical self-similarity over a range of timescales) [6-8], techniques that can characterise complex dynamics invisible to conventional linear signal processing tools, provide indicators that are successfully able to separate normal from disordered subjects. Simple machine learning techniques can then discriminate between normal and disordered subjects to better than 90% accuracy [4].
Simple Measures for Symptom Severity Monitoring
Using a principled Bayesian approach, the project aims to systematically identify the best combinations of existing and novel signal analysis methods for detecting the presence and extent of disorder, to assist objective diagnosis and monitoring of the clinical progress of patients. The figure demonstrates the separation using support vector machines,
of normal or asymptomatic (black dots) from disordered (blue crosses) patients using recurrence probability density entropy and a novel, psychoacoustically-informed measure - pitch period entropy. It also shows
the kernel probability density of these two measures (bottom left) projected down onto the first
linear discriminant analysis co-ordinate, to provide a simple, single scale measure of symptom severity for continuous, treatment effectiveness monitoring.
For further details, please contact Max Little.
[1] S. v. Campenhausen, B. Bornscheina, R. Wick, K. Bötzel, C. Sampaio, W. Poewee, W. Oertel, U. Siebert, K. Berger, and R. Dodel (2005), Prevalence and incidence of Parkinson's disease in Europe. European Neuropsychopharmacology, 15:473-490.
[2] B. Harnett (2006), Telemedicine systems and telecommunications. J Telemed Telecare, 12:4-15.
[3] M. Little, P. McSharry, I. Moroz, and S. Roberts (2006), Testing the assumptions of linear prediction analysis in normal vowels. Journal of the Acoustical Society of America, 119:549-558.
[4] M. Little, P. McSharry, I. Moroz, and S. Roberts (2006), Nonlinear, biophysically-informed speech pathology detection, in Proc ICASSP 2006. New York: IEEE Publishers.
[5] M. A. Little, P. E. McSharry, D. A. E. Costello, S. J. Roberts, and I. M. Moroz (2007), Exploiting Recurrence and Fractal Scaling Properties for Voice Disorder Detection, Biomedical Engineering Online, 6.
[6] M. D. LaMar, Y. Y. Qi, and J. Xin (2003), Modeling vocal fold motion with a hydrodynamic semicontinuum model. Journal of the Acoustical Society of America, 114:455-464.
[7] M. H. Krane (2005), Aeroacoustic production of low-frequency unvoiced speech sounds. Journal of the Acoustical Society of America, 118:410-427.
[8] M. Little, P. McSharry, I. Moroz, and S. Roberts (2005), A Simple, Quasi-linear, Discrete Model of Vocal Fold Dynamics, in Nonlinear Analyses and Algorithms for Speech Processing, pp. 348-356.