Digital Speech Assessment in Parkinson's Disease

Bringing Objective Speech Measurement Into Parkinson’s Disease Research

Speech is among the earliest functions affected in Parkinson’s disease, and digital speech assessment is one of the few ways to measure it objectively and often. Reduced loudness, monotone delivery and imprecise articulation appear early and persist across the disease course. Yet the standard clinical instrument, the MDS-UPDRS, captures speech in just two brief subjective items, scored episodically at clinic visits.

For drug developers, that leaves one of the most accessible symptom domains in Parkinson’s largely unmeasured between visits. Smartphones can record speech anywhere, and modern analytics can quantify it. The open question has been simpler than the technology. Can a remote protocol deliver usable data from a Parkinson’s population, week after week, at a quality that supports analysis?

Koneksa set out to answer that question. Participants across three cohorts completed weekly at-home speech assessments over nine weeks, benchmarked against clinician assessments.

Why episodic speech assessment leaves the signal on the table

Two subjective items, rated a handful of times over a trial, cannot characterize a symptom domain that varies day to day. Clinician-rated speech scoring is also coarse by design, and it cannot detect change below the threshold of human perception.

The result is a measurement gap. Sponsors developing therapies in Parkinson’s have no objective, quantitative, remotely deployable measure of vocal impairment. They also have no way to observe it in the environments where patients actually speak.

A remote, multi-measure speech strategy

Koneksa designed and coordinated a prospective, multi-cohort observational study spanning the Parkinson’s risk continuum: participants with Parkinson’s disease, participants in the prodromal phase, and healthy controls.

Provisioned smartphones and weekly at-home speech assessments across nine weeks
Three speech tasks, sustained phonation, sentence reading and repeated syllables, generating 30+ quantitative measures
Third-party clinical-grade speech analytics integrated into the Koneksa data pipeline rather than run as a standalone tool
Virtual MDS-UPDRS assessments at baseline, week 4 and week 8, with all digital measures benchmarked against MDS-UPDRS and Hoehn & Yahr staging
Device provisioning, participant coordination, compliance monitoring and data export managed on one configurable platform

Download the Full Case Study

Complete the form to access the full resource.

X/Twitter

This field is for validation purposes and should be left unchanged.

Name(Required)

Email(Required)

Phone Number

Company Name(Required)

Job Title(Required)

What You'll Learn

How 80+ participants across three cohorts completed weekly at-home speech assessments over nine weeks, and what compliance and data quality looked like in practice

Which speech measures reached endpoint-grade test-retest reliability, and which tracked clinician-rated symptom severity

Why sex, not disease status, was the single largest source of variance in the voice data, and what that means for how you size and stratify a speech study

Where transfer learning from automatic speech recognition models fits, and what it still requires before it can be relied on

How the study was structured across Koneksa, an academic site and a speech analytics partner, and what that division of responsibility made possible