Speech biomarkers for DM1 and neuromuscular trials
Cephalgo gives DM1, late-onset Pompe disease and SMA trials more frequent readings of articulation, phonation and respiratory support from speech, in recordings you hold, site visits or home, stable across phone and software changes. Device changes ›
Bulbar and respiratory weakness is measured only at visits
In DM1, late-onset Pompe disease and SMA, the signs that speech carries are measured rarely.
Mostly at visits
Bulbar weakness is rated mostly at site visits, with little in between.
Small, varied cohorts
Rare-disease cohorts are small and span ages and severity, so every reading per participant counts.
Patient reports miss change
In DM1, reduced awareness of symptoms weakens patient-reported measures, so slow bulbar decline goes unreported.
Where speech adds a measure
Bulbar and respiratory weakness show in speech. Site-visit recordings and a few minutes at home add readings between visits, as an exploratory or secondary endpoint.
Myotonic dystrophy type 1
Dysarthria, hypernasal speech and oral myotonia, including the warm-up effect across repeated syllables. Read beside the communication subscale of the Myotonic Dystrophy Health Index (MDHI).
Late-onset Pompe disease
Phonation and respiratory support recorded upright and supine, read beside the site’s forced vital capacity (FVC) in the same positions.
Spinal muscular atrophy
Bulbar function in SMA, with tasks set per age and SMA type; for children, parental consent and assent as the protocol specifies. Read beside the bulbar items your protocol scores.
Joined to the visit score
Readings joined to the visit score, one line per participant. In dose‑finding studies, per‑arm readings go only to the independent unblinded statistician and data monitoring committee.

Validated in your study, against your reference
We compare our readings with the clinician-rated, patient-reported or measured reference your trial already runs.
Held-out participants
We evaluate on held‑out participants from your study, never used for calibration.
No black box
Each reading comes with the named speech features behind it and its analysis version, so any reading can be traced and rerun. Our pipeline ›
Small datasets
Our already-trained model is calibrated on screening and baseline data, blind to arm, locked before any post‑baseline reading. What goes into the analysis plan ›
- DM1
- MDHI communication subscale
- Late-onset Pompe
- FVC upright and supine
- SMA
- Bulbar items your protocol scores
Re‑analysis of the DM1 and neuromuscular trials you already ran
Your past trials hold audio never read for speech. You get readings on the reference your trial used, from a completed or ongoing trial where consent covers further analysis. No new collection, no site burden, a fixed fee.
On your scale
You get a reading per recording, joined to the visit score, calibrated on screening and baseline ratings and locked blind to arm before post‑baseline recordings are read, so no post‑baseline data shapes the model.
For your next trial
You get variance components and, from a completed trial, an effect‑size estimate your statistician computes on the locked readings, so a small, varied cohort is sized on real data.
Joined to your reference
Readings joined to the MDHI communication subscale, FVC or bulbar items your trial recorded, per participant and visit.
Named features
You get the articulation, phonation and respiratory features behind every reading, so your team sees what moved.
- You send
- Recordings or an authorization to their holder (we draft it), visit scores and the consent scope
- Readings
- The reference your trial used, per recording
- Planning
- Effect size and variance for the next trial
- Reference
- MDHI, FVC or bulbar items, joined per visit
- New collection
- None
- Fee and timeline
- Fixed in writing before any data moves
How a session runs
Readings from audio you already hold, site-visit recordings, and home sessions via your eCOA or app vendor or the participant’s own phone.
Short spoken tasks
Sustained vowels, syllable repetition, reading or picture description, set per cohort and age, and read the same way in multilingual studies.
With the capture your study uses
Your eCOA or app vendor, or the participant’s own phone, with quality flags on every recording. Prospective study ›
The same reading after a phone update
When a participant’s phone updates or is replaced, you get readings recalibrated from the audio alone, blind to arm, so readings before and after the change sit on one scale. Device changes ›
Check speech-related entry criteria before the screening visit
A short recording on the candidate’s own phone. Diagnosis, genetic confirmation, FVC and other measured criteria stay with the site.
Your entry criteria
Speech-related criteria in your protocol, such as a bulbar or speech score, read against an entry range fixed in writing before the first candidate.
Cephalgo in neuromuscular trials
Which neuromuscular diseases does Cephalgo cover?
Cephalgo covers myotonic dystrophy type 1 (DM1), late-onset Pompe disease and spinal muscular atrophy (SMA), where bulbar and respiratory weakness show in speech.
Does Cephalgo work with small rare-disease cohorts?
Yes. Cephalgo’s already-trained model is calibrated on your study’s screening and baseline data, blind to arm, locked before any post‑baseline reading, and evaluated on held‑out participants.
Do phone and software changes affect speech biomarker readings in neuromuscular trials?
Phone and software changes shift speech measures. Cephalgo readings stay stable across them, recalibrated from the audio, so a change in articulation, phonation or respiratory support comes from the participant. Device changes ›
What are Cephalgo’s speech readings compared with?
Cephalgo compares its readings with your trial’s own reference: the MDHI communication subscale in DM1, upright and supine forced vital capacity in late-onset Pompe, and protocol bulbar items in SMA.
