Glottal Source Analysis of Voice Deficits in Newly Diagnosed Drug-naïve Patients with Parkinson's Disease: Correlation Between Acoustic Speech Characteristics and Non-Speech Motor Performance. (March 2020)

• Record Type: Journal Article
• Authors: Novotný, Michal; Dušek, Petr; Daly, Imen; Růžička, Evžen; Rusz, Jan
• Is Part Of: Biomedical signal processing and control. Volume 57(2020)
• Publication Date: 2020-03
• Subjects: Glottal source -- perturbation -- cepstral peak prominence -- phonation -- Parkinson's disease; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.101818
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Perturbation measures, cepstral peak prominence, and glottal source analysis were compared. The glottal source assessment provided strongest marker of dysphonia in Parkinson's disease. The classification experiment based on phonatory markers reached AUC = 0.78 The harmonic richness factor showed relationship with the rigidity and bradykinesia. Phonatory assessment is useful but not self-sufficient for classification of Parkinson's disease. Abstract: As dysphonia is assumed to be the most prevalent feature of hypokinetic dysarthria in Parkinson's disease (PD), several acoustic approaches have been introduced for its assessment. However, the sensitivity of different acoustic measures to the occurrence of dysphonia in drug-naïve de-novo patients with PD has not yet been explored. The goal of this study is to examine the performance of glottal source parametrization and compare it with traditional perturbation and cepstral measures. Sustained phonations of 40 de-novo drug-naïve PD participants and 40 age- and sex-matched healthy controls were recorded. Each utterance was ranked by a set of glottal source parameters obtained by inverse adaptive filtering, the perturbation parameters, and cepstral peak prominence measures. Our results revealed the significant differences in the glottal source parameter Harmonic Richness Factor (HRF: p < 0.01) and cepstral peak prominence parameter ( p < 0.05). The support vector machine classification between PD and healthy controls reached the area under the curve of 0.78. The analysis of relationships showed significant negative correlations between HRF and total Movement Disorders Society – Unified Parkinson's Disease Rating Scale III score ( r =-0.35, p < 0.05), as well as its rigidity ( r =-0.43, p < 0.01) and bradykinesia ( r =-0.32, p < 0.05) sub-scores. Glottal source assessment appears to be a superior method to assess PD-related dysphonia compared to the traditional perturbation and cepstral approaches. Our results highlight that dysphonia and limb bradykinesia and rigidity in PD are controlled by similar underlying brain processes.