Biomedical speech signal insights from a large scale cohort across seven countries: The Parkinson’s voice initiative study

Biomedical speech signal insights from a large scale cohort across seven countries: The Parkinson’s voice initiative study. Tsanas, A. & Arora, S. In Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA), pages 45-48, 2019.

Paper abstract bibtex

Previous work has demonstrated the enormous potential of speech signals collected under highly controlled acoustic conditions in biomedical applications. These include accurately differentiating people diagnosed with Parkinson’s Disease (PD) from Healthy Controls (HC), and longitudinal telemonitoring of PD symptom severity. The generalizability and scalability of these findings need to be investigated when speech signals are not recorded under optimal, carefully controlled acoustic conditions. In this regard, we recently completed the Parkinson’s Voice Initiative (PVI) study collecting data from a very large cohort comprising 1483 PD and 8300 HC participants from seven countries. Specifically, we collected 19,303 sustained vowel /a/ recordings: 144 (Argentina), 227 (Brazil), 1521 (Canada), 75 (Mexico), 573 (Spain), 4088 (UK) and 12,675 (USA). We acoustically characterized these recordings using 307 dysphonia measures which we had previously investigated in related PD studies. We draw comparisons against previous studies which processed high quality speech data, and their generalizability in this large-scale cohort. We found that many of the state-of-art nonlinear dysphonia measures do not differentiate PD and HC sufficiently well, likely because of the reduced signal bandwidth. These exploratory findings provide new insights into understanding the challenges in the PVI dataset, underlining the need for further speech signal processing development.

@inproceedings{
 title = {Biomedical speech signal insights from a large scale cohort across seven countries: The Parkinson’s voice initiative study},
 type = {inproceedings},
 year = {2019},
 keywords = {Parkinson’s Disease (PD),Parkinson’s Voice Initiative (PVI),Speech signal processing,Sustained vowel phonations},
 pages = {45-48},
 city = {Florence, Italy},
 id = {ecec1d90-fe39-335a-aec9-e59a5adbd80c},
 created = {2021-08-11T17:28:52.460Z},
 file_attached = {true},
 profile_id = {63c14ecf-8d9b-3014-983a-e77627cb99e3},
 last_modified = {2021-08-11T17:29:58.610Z},
 read = {false},
 starred = {false},
 authored = {true},
 confirmed = {true},
 hidden = {false},
 private_publication = {false},
 abstract = {Previous work has demonstrated the enormous potential of speech signals collected under highly controlled acoustic conditions in biomedical applications. These include accurately differentiating people diagnosed with Parkinson’s Disease (PD) from Healthy Controls (HC), and longitudinal telemonitoring of PD symptom severity. The generalizability and scalability of these findings need to be investigated when speech signals are not recorded under optimal, carefully controlled acoustic conditions. In this regard, we recently completed the Parkinson’s Voice Initiative (PVI) study collecting data from a very large cohort comprising 1483 PD and 8300 HC participants from seven countries. Specifically, we collected 19,303 sustained vowel /a/ recordings: 144 (Argentina), 227 (Brazil), 1521 (Canada), 75 (Mexico), 573 (Spain), 4088 (UK) and 12,675 (USA). We acoustically characterized these recordings using 307 dysphonia measures which we had previously investigated in related PD studies. We draw comparisons against previous studies which processed high quality speech data, and their generalizability in this large-scale cohort. We found that many of the state-of-art nonlinear dysphonia measures do not differentiate PD and HC sufficiently well, likely because of the reduced signal bandwidth. These exploratory findings provide new insights into understanding the challenges in the PVI dataset, underlining the need for further speech signal processing development.},
 bibtype = {inproceedings},
 author = {Tsanas, Athanasios and Arora, Siddharth},
 booktitle = {Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA)}
}

Downloads: 0