Parkinson’s disease
Speech biomarkers
for Parkinson's trials
A short recording on the participant’s own phone is read on the MDS‑UPDRS speech item. Recordings the study already holds can be read the same way.
Where the reading lands
From the MDS‑UPDRS to what Cephalgo reads
Cephalgo's reading lands on the MDS‑UPDRS speech item.
Input
A couple of minutes of speechOn the participant's own phone, between visits
Analysis
Three streams, one multi-stream modelDeep-learning embeddings · acoustic-linguistic · cognitive-linguistic
Output · MDS‑UPDRS speech item
0Normal
1Slight
2Mild
3Moderate
4Severe
Each reading lands on the item's 0 to 4 grade.
In practice
Short spoken tasks, read on the MDS‑UPDRS speech item
The tasks
Customised to your cohort
For example: sustained vowels, rapid syllable repetition and reading aloud. Tasks and their number are set for each cohort.
Where
At home or in clinic
On the participant's own device, with an automatic quality check at capture.
The output
Between scheduled visits
A reading on the MDS‑UPDRS speech item at every session, followed between visits.
In the trial
Built for real‑world clinical trials
Objective and automated
The same scoring everywhere
Every site and every visit is scored the same way, with no rater in the loop.
On the trial's scale
Readings on the MDS‑UPDRS speech item
Each reading lands on the scale the protocol already carries.
Robust
Limits drift when processing changes
Recalibrated from the recordings when on-device processing changes. How it works
Patient-centric
A couple of minutes at home
On the participant's own phone, tablet or computer, between scheduled visits.
Transparent and trackable
Visible per session
Completion, recording quality and versions are logged with every session.
Integrated
No new visit, no new rater
Runs on the Cephalgo platform alongside the assessments the study already runs.
Evidence
Validation following DiMe V3
Following the Digital Medicine Society’s V3 framework.
Clinical validation
MDS‑UPDRS speech item
The reading was evaluated in a held‑out smartphone cohort, and Parkinson's separated from controls in an independent clinical cohort.
Analytical validation
Processing change
Robustness to on-device processing change, presented at the Digital Biomarkers in Clinical Trials Summit in Basel and at ISCTM, 2026.
Publications
Published and presented
White paper
Speech-Based Assessment in Parkinson's Disease
Request the white paper
ePoster · AD/PD 2026
Validation of fully explainable speech biomarkers for PD/HC screening
Parkinson's against healthy controls.
Basel and ISCTM · 2026
When the Device Changes Mid-Trial
Robustness to on-device processing variability.
Contact
Where would speech sit in your next PD protocol?
Tell us about the studies you’re running and the questions you want speech to answer.
Get in touch →