html5-img
1 / 1

Accurate Telemonitoring of parkinson’s disease Symptom severity using speech signals

 Parkinson’s patient speaks into microphone.  Predicted UPDRS report to clinical staff.  Home telemonitoring device records speech signal.  Statistical mapping of algorithms to UPDRS.  Speech transferred to USB stick. Speech signal processing algorithms.

tieve
Download Presentation

Accurate Telemonitoring of parkinson’s disease Symptom severity using speech signals

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1.  Parkinson’s patient speaks into microphone Predicted UPDRS report to clinical staff  Home telemonitoring device records speech signal Statistical mapping of algorithms to UPDRS  Speech transferred to USB stick • Speech • signal • processing algorithms J=|F0,i-F0,i+1| S=|A0,i-A0,i+1| Internet  Data into patient’s personal computer  Data into dedicated server in the clinic Patient’s home Medical Centre Accurate Telemonitoring of parkinson’s disease Symptom severity using speech signals Athanasios Tsanas1,2, Max A. Little1,2,3, Patrick E. McSharry1,2, Lorraine O. Ramig4,5 1Systems Analysis, Modelling and Prediction (SAMP), Mathematical Institute, University of Oxford, Oxford, UK, 2Oxford Centre for Industrial and Applied Mathematics (OCIAM), University of Oxford, Oxford, UK,3Oxford Centre for Integrative Systems Biology, Department of Physics, University of Oxford, UK, 4Speech, Language, and Hearing Science, University of Colorado, Boulder, Colorado, USA, 5National Center for Voice and Speech, Denver, Colorado, USA Features – various approaches Parkinson’s disease symptom tracking Project background • Parkinson’s disease (PD) claims lives at an epidemic rate (affecting ~20/100,000 people every year) • There is no treatment, but drugs can alleviate some of the symptoms • Clinical metric used to quantify average symptom severity: Unified Parkinson’s Disease Rating Scale (UPDRS). Motor-UPDRS range: 0-108, Total-UPDRS range: 0-176, where 0 denotes healthy control. • Currently, UPDRS is estimated by clinical raters (subjective, inter-rater variability) • PD affects speech, and there is empirical evidence of degrading speech performance with disease progression • We propose novel nonlinear signal processing algorithms mining the information in speech • We demonstrate that by using speech signals we can accurately replicate the clinicians’ UPDRS assessment Feature selection • Curse of dimensionality (failure to adequately populate the feature space) • Reducing number of features enables a) improved performance, b) more accurate inference of the underlying characteristics of the modelled system • Many approaches: LASSO, elastic net, Random Forests (work in progress) Conclusions Schematic representationof the UPDRS estimation process • Speech signals convey clinically useful information • Fast, accurate, remote, objective monitoring of Parkinson’s disease is shown to be possible using simply speech signals • Results are considerably better than the inter-rater variability (which is about 5 UPDRS points). • This technology could facilitate large-scale clinical trials into novel Parkinson’s disease treatments Statistical mapping • Statistical machine learning algorithm maps the feature matrix on the response • Experimented with various state of the art classification and regression algorithms • Random Forests seems to work particularly well in this problem, probably because there are many features contributing towards the response variable Results References • A. Tsanas, M.A. Little, P.E. McSharry, L.O. Ramig: “Accurate telemonitoring of Parkinson’s disease progression by non-invasive speech tests”, IEEE Transactions Biomedical Engineering, Vol. 57, pp. 884-893, 2010a • A. Tsanas, M.A. Little, P.E. McSharry, L.O. Ramig: “Enhanced classical dysphonia measures and sparse regression for telemonitoring of Parkinson’s disease progression”, IEEE Signal Processing Society, International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 594-597, Dallas, Texas, US, 14-19 March 2010b • A. Tsanas, M.A. Little, P.E. McSharry, L.O. Ramig: “New nonlinear markers and insights into speech signal degradation for effective tracking of Parkinson’s disease symptom severity", International Symposium on Nonlinear Theory and its Applications (NOLTA), pp. 457-460, Krakow, Poland, 5-8 September 2010c (invited) • A. Tsanas, M.A. Little, P.E. McSharry, L.O. Ramig: “Remote tracking of Parkinson’s disease progression by extracting novel dysphonia patterns from speech signals”, Journal of the Royal Society Interface, (in press) 2010d • Use 10-fold cross validation with 100 repetitions for statistical confidence • Report the out of sample mean absolute error (MAE). The findings are given in the form mean ± standard deviation • Use the one-standard-error rule to determine the most parsimonious model

More Related