|
Descripción
|
|
|---|---|
| Idiopathic Parkinson's disease (PD) is a chronic degenerative disorder affecting the dopamine production centers in the basal ganglia and which is mainly manifested with dysfunctions in motor systems. The disease affects 2% of the population over 60 years but its prevalence is likely to increase due to the aging trend of the population. Due to the overlapping of the clinical signs with other atypical parkinsonisms, the diagnosis usually takes around 3 years, limiting the effect and prognosis of the treatments, and consequently the quality of life and life expectancy of the patients. Thus, early diagnosis and treatment are vital to alleviate the effects of the disease. However, despite of the efforts, to date there are not early and non-invasive robust markers of the disease. Some of these efforts are focused in the analysis of certain biometric patterns such as gait, speech, handwriting, and typing patterns. The automatic analysis of the speech is an emerging field with potential to generate quantitative non-invasive biomarkers derived from clinical tests, and pioneering studies are identifying that the speech is affected even in the pre- clinical phase. Thus, our goal is to identify reliable biomarkers based on digital signal processing and artificial intelligence techniques extracted from the speech. This would open the possibility to design non-invasive automatic detection and screening systems, enabling a change of paradigm in early PD monitoring and in its differential evaluation with respect to other atypical parkinsonisms. In this context, our focus is in the identification of biomarkers directly or indirectly linked with articulatory and kinetic aspects of the speech, since they are supposed to be highly correlated with the existing motor symptoms. Our preliminary data and experiments show the potentiality of the artificial intelligence techniques to identify the disease and to assess its extent. Our conclusions are based on a prospective study with different corpora of parkinsonian and control subjects modelled following different approaches, mainly grounded on generative models of the speech, its phonetic units, and its manners of articulation. Results obtained are consistent among the different approaches and corpora, even in cross database scenarios, providing detection accuracies over 85%. This talk will present the approaches we followed and the results obtained, and will discuss some key aspects for the future research in the field. | |
|
Internacional
|
Si |
|
Nombre congreso
|
1st Automatic Assessment of Parkinsonian Speech Workshop, |
|
Tipo de participación
|
730 |
|
Lugar del congreso
|
Cambridge, Masachussetts, EEUU |
|
Revisores
|
Si |
|
ISBN o ISSN
|
|
|
DOI
|
|
|
Fecha inicio congreso
|
20/09/2019 |
|
Fecha fin congreso
|
21/09/2019 |
|
Desde la página
|
6 |
|
Hasta la página
|
6 |
|
Título de las actas
|
1st Automatic Assessment of Parkinsonian Speech Workshop |