Skip to Main content Skip to Navigation

Détection précoce de la maladie de Parkinson par l'analyse de la voix et corrélations avec la neuroimagerie

Laetitia Jeancolas 1, 2, 3
SAMOVAR - Services répartis, Architectures, MOdélisation, Validation, Administration des Réseaux
Abstract : Vocal impairments, known as hypokinetic dysarthria, are one of the first symptoms to appear in Parkinson's Disease (PD). A large number of articles exist on PD detection through voice analysis, but few have focused on the early stages of the disease. Furthermore, to our knowledge, no study had been published on remote PD detection via speech transmitted through the telephone channel. The aim of this PhD work was to study vocal changes in PD at early and preclinical stages, and develop automatic detection and monitoring models. The long-term purpose is to build a cheap early diagnosis and monitoring tool, that doctors could use at their office, and even more interestingly, that could be used remotely with any telephone. The first step was to build a large voice database with more than 200 French speakers, including early PD patients, healthy controls and idiopathic Rapid eye movement sleep Behavior Disorder (iRBD) subjects, who can be considered at PD preclinical stage. All these subjects performed different vocal tasks and were recorded with a professional microphone and with the internal microphone of a computer. Moreover, they called once a month an interactive voice server, with their own phone. We studied the effect of microphone quality, speech tasks, gender, and classification analysis methodologies. We analyzed the vocal recordings with three different analysis methods, covering different time scale analyses. We started with cepstral coefficients and Gaussian Mixture Models (GMM). Then we adapted x-vectors methodology (which never had been used in PD detection) and finally we extracted global features classified with Support Vector Machine (SVM). We detected vocal impairments at PD early and preclinical stages in articulation, prosody, speech flow and rhythmic abilities. With the professional microphone recordings, we obtained an accuracy (Acc) of 89% for male early PD detection, just using 6min of reading, free speech, fast and slow syllable repetitions. As for women, we reached Acc = 70% with 1min of free speech. With the telephone recordings, we achieved Acc = 75% for men, with 5min of rapid syllable repetitions, and 67% for women, with 5min of free speech. These results are an important first step towards early PD telediagnosis. We also studied correlations with neuroimaging, and we were able to linearly predict DatScan and Magnetic Resonance Imaging (MRI) neuromelanin sensitive data, from a set of vocal features, in a significant way. This latter result is promising regarding the possible future use of voice for early PD monitoring.
Complete list of metadata

Cited literature [180 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Friday, February 7, 2020 - 3:02:13 PM
Last modification on : Monday, August 24, 2020 - 4:16:13 PM
Long-term archiving on: : Friday, May 8, 2020 - 4:34:37 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02470759, version 1


Laetitia Jeancolas. Détection précoce de la maladie de Parkinson par l'analyse de la voix et corrélations avec la neuroimagerie. Traitement du signal et de l'image [eess.SP]. Université Paris-Saclay, 2019. Français. ⟨NNT : 2019SACLL019⟩. ⟨tel-02470759⟩



Record views


Files downloads