Tabea Thies, Felix Dörr, Jan Rusz, Johannes Tröger
*Presented at the 30th World Congress on Parkinson’s Disease and Related Disorders, May 2025
Abstract
Background: Speech impairments in neurodegenerative diseases include deficits in articulation and phonation reducing speech intelligibility. This study evaluates an automatic speech analysis method to characterize speech performance across isolated REM sleep behavior disorder (iRBD), Parkinson’s disease (PD), and progressive supranuclear palsy (PSP) using intelligibility scores and acoustic features related to phonation and articulation. Speech performance is compared to healthy controls (HC).
Methods: Clinical and speech data were collected of a Czech cohort: Speech Assessment: Reading, Sustained phonation of the vowel “A”, Rapid syllable repetition (/pataka/). Motor Assessment: PSP: Natural History and Neuroprotection in Parkinson Plus Syndromes (NNIPPS), iRBD & PD: Motor part III of the Unified Parkinson’s Disease Rating Scale (UPDRS III)
Data Processing: ki:elements’ signal processing pipeline SIGMA was used to extract ki: SB-M scores related to intelligibility, phonation and articulation from the raw audio files. Group differences were analyzed using Mann-Whitney U tests, with p-values adjusted using the Benjamini-Hochberg method.
Results: Intelligibility is comparable between HC and iRBD but progressively declines in PD and PSP. Articulatory deficits are evident in iRBD and become more pronounced in PD and PSP. Phonatory deficits are exclusively observed in PSP.
Conclusion: Articulatory changes are already detectable in iRBD without impacting intelligibility. There is a progressive deterioration of speech functions from iRBD to PD to PSP. Automatic speech analysis is capable to detect these pattern, paving the way for more time-efficient analytical approaches.
