Low-burden digital speech biomarkers for early diagnosis and prognosis in Parkinson’s Disease: Study Design and Protocol

Alexander Jäck1,2, Ebru Baykara4, Mira Fischer1,2, Urban M. Fietzek1,2, Alexander Bernhardt1,2, Alexandra König4, Johannes Tröger4, Günter Höglinger1,2,3, Johannes Levin1,2,3

1) Department of Neurology, University Hospital of Munich, LMU Munich, Munich, Germany. 2) German Center for Neurodegenerative Diseases (DZNE), Munich, Germany. 3) Munich Cluster for Systems Neurology (SyNergy), Munich, Germany. 4) ki:elements GmbH, Saarbrücken, Germany.

* Poster presented at the AD/PD™ Alzheimer’s disease and Parkinson’s disease Conference 2024, Portugal



Diagnosing Parkinson’s Disease (PD), the leading neurodegenerative movement disorder, predominantly relies on clinical motor-function criteria. This method, while sidestepping costly diagnostics, is not without challenges. Its effectiveness is influenced by the neurologist’s expertise, with diagnostic precision often not surpassing 80%. Non-motor symptoms, such as autonomic dysfunction and REM Sleep Behavior Disorder (RBD), frequently manifest prior to motor symptoms, suggesting their potential as early biomarkers for prodromal PD. Furthermore, there are no reliable biomarkers for monitoring and predicting clinical outcomes of established PD. Recent studies have highlighted speech and linguistic alterations in neurodegenerative disorders like PD as promising biomarkers to address these diagnostic challenges. The potential of leveraging speech patterns in neurodegenerative disease, has prompted research into artificial intelligence (AI)-driven diagnostic tools. AI-powered mobile apps, democratize healthcare access, especially in remote settings, offering safe and timely patient care. The integration of speech analysis via AI-powered remote assessments could thus offer a particularly valuable approach to addressing some current challenges in PD.


In collaboration with ki:elements, this study seeks to identify and validate digital speech biomarkers for PD through AI empowered remote assessments. Building on ki:elements’ prior work, the focus is on detecting early-stage Parkinson’s, tracking progression, and assessing levodopa responsiveness. A subproject will be further validate digital speech biomarkers against established molecular biomarkers including Tau, Amyloid and lewy-fold a-Synuclein.


The study will commence as a 12-month observational research, aiming to include 100 patients and 100 healthy controls, building on established cohorts with multiparametric deep phenotyping established at LMU Munich. In addition to extensive clinical phenotyping, this study will collect comprehensive speech data remotely through ki:element’s Mili mobile app. Combined with gold standard measures and advanced speech analysis using AI, this data will help to investigate speech and language based phenotypes for PD patients, capturing motor, cognitive, and mood symptoms.


The intricate diagnostic landscape of PD is riddled with challenges due to its reliance on clinical motor-function criteria and poor prognostic assessments. The emergence of non-motor symptoms as precursors to motor impairments showcases the complexity of PD’s clinical manifestation. Considering the precursory and progressive nature of speech alterations AI and machine learning also offer immense promise in establishing reliable digital biomarkers. By delving into both speech alterations and molecular biomarkers in an exploratory subproject we seek to shed light on the underlying pathomechanisms of speech and linguistic changes in PD.

Share this article