Objective: To investigate speech changes induced by subthalamic deep brain stimulation (DBS) in persons with Parkinson’s disease (pwPD).

Background: Dysarthria is a common debilitating motor symptom of PD with major consequences for patient autonomy and quality of life. Effect of subthalamic DBS on speech performance can be variable. While some studies reported improvement, DBS-induced dysarthria is a well-known phenomenon that may hamper therapeutic success.

Method: We collected speech recordings of 56 German native speaking pwPD after overnight withdrawal of dopaminergic medications. Four speech tasks (vowel phonation, oral diadochokinesis, reading, and storytelling) were performed in DBS-OFF and DBS-ON condition. All recordings were acquired in a soundproof booth with a condenser microphone headset. Automatic speech analysis quantified 19 established acoustic speech features. Four unweighted composite scores were derived, characterizing speech subsystems (articulation, phonation/respiration, temporal dynamics and voice quality). Two human raters evaluated whether any DBS-induced speech alterations were perceptible by comparing OFF and ON recordings.

Linear mixed modelling identified speech features with a significant DBS effect. A Random Forest Model (RFM) determined the discriminative value of the assessed speech features. A second RFM quantified the relevance of each feature for perceivable changes. K-Means clustering explored distinct DBS response patterns, characterized with a third RFM.

Results: We observed significant DBS-induced speech alterations (9/19 features). All features improved or were not affected. Single feature analysis and composite scores demonstrated prominent DBS-induced changes of respiration and phonation, confirmed by the corresponding RFM [figure1]. The perceptual salience of speech alterations was predominantly driven by loudness [figure2]. Clustering analysis discerned two distinct DBS response patterns [figure3; figure4]. Cluster 2 exhibited larger DBS related changes in the discriminative features.

Conclusion: This investigation demonstrates significant DBS-induced improvements of multiple speech domains in pwPD, mostly related to respiration, phonation and increased speech modulation. Exploratory analysis suggests effect sizes are interrelated and driven by a common mechanism. We will explore potential neuroanatomical correlates based on patient-specific stimulation simulations and voxel-wise statistics.

Relative Feature Importance DBS-OFF vs DBS-ON

Relative Feature Importance for Speech Perception

Principal Component Analysis (PCA)

Relative Feature Importance for Cluster Analysis

References: [1] Hlavnička, Jan. “Automated Analysis of speech disorders in neurodegenerative diseases.” Doctoral Thesis, Dept. of Circuit Theory, Faculty of Electrical Engineering, Czech Technical University in Prague. 2018.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/characterization-of-dbs-induced-speech-improvements-in-a-german-pd-cohort/