*Presented at MDS Congress 2026

Objective

To evaluate the feasibility and clinical validity of remote, digital speech-based cognitive and speech motor assessments in Parkinson’s disease (PD).

Background

Speech in PD reflects both motor and cognitive impairment, offering a scalable source for digital biomarkers. While speech tasks can be collected remotely, evidence is needed that app-based assessments are feasible at home and yield measures that relate meaningfully to established clinical outcomes.

Methods

24 German-speaking individuals with PD were enrolled from the DESCRIBE study in Munich [table1]. Diagnosis followed the established MDS criteria [1]. Clinical assessment included part III of the Unified Parkinson’s Disease Rating Scale (total score, speech item), and the Montreal Cognitive Assessment (MoCA).

Participants completed one home-based session using the Mili mobile app, including speech-based cognitive tests (verbal fluency; verbal learning) and speech motor tasks (maximum vowel phonation, reading, free speech, oral diadochokinesis “pataka”). Speech samples were processed using SIGMA, ki:elements proprietary processing pipeline to derive speech-based cognitive (SB-C) total score and subscores for memory, processing speed, and executive functioning [2], and speech-based motor (SB-M) Intelligibility Score [3]. Additionally, selected SB-M single features were extracted and will be analyzed for conference presentation. Clinical validation involved Spearman correlations between SB-C and MoCA, and SB-M with UPDRS III and disease duration.

Results

Participant characteristics and averaged performance are reported [table 1, table 2]. SB-C total score correlated with MoCA (ρ=.496, p=.016). SB-C subscores showed significant associations for processing speed (ρ=.508, p=.013) and memory (ρ=.438, p=.037), and a trend for executive functions (ρ=.366, p=.086). SB-M Intelligibility correlated with UPDRS III total (ρ=−.475, p=.019) and showed trends with the UPDRS III speech item (ρ=−.371, p=.074), and disease duration (ρ=−.367, p=.077).

Conclusions

Remote, app-based speech assessment in PD is feasible and yields automated measures that align with clinical gold standards for motor and cognitive function. These findings support remotely acquired speech as a scalable digital biomarker approach for identification and monitoring PD severity in real-world settings.