Three million voice recordings sit behind an algorithm the FDA now treats as a potential breakthrough for the deadliest chronic condition in medicine. Noah Labs Vox reads a five-second daily voice clip for the acoustic fingerprints of a failing heart, and its maker just cleared a regulatory gate that most diagnostics never reach.
A 5-second daily voice sample can flag worsening heart failure up to 21 days before a hospitalization event.
Vocal biomarkers are crossing from research to commercial diagnostics, but the field's first casualty is already on the record.
The heart speaks before it breaks
Heart failure is the leading cause of hospitalization among adults worldwide, with more than one million admissions every year in the United States alone. About 6.7 million Americans live with the condition, and that number is projected to climb to 8.7 million by 2030.
Noah Labs Vox training dataset
The algorithm is built on what Noah Labs calls the world's largest exclusive dataset of heart-failure patient voice recordings, validated across five multicenter trials. · Noah Labs, 2026
The monitoring that exists today works, but it is invasive or it is late. The dominant remote-monitoring tool, Abbott's CardioMEMS, is a sensor implanted into the pulmonary artery that reads blood pressure from inside the heart. It gained FDA clearance in 2014 and works well enough that Britain's NICE recommended it for the National Health Service in March 2026.
Implanting a sensor into a failing heart is not something a general practitioner orders for a patient in a rural clinic. Noah Labs Vox takes the opposite route: no hardware at all. The patient records a five-second clip on a phone, and the model reads the vocal folds for the swelling that accompanies pulmonary congestion and fluid overload.
How a voice clip becomes a cardiac signal
The algorithm is software-only. It processes acoustic features from a standard smartphone microphone, compares them against millions of labelled recordings, and flags the early signs of decompensation up to 21 days before a hospitalization event.
Voice is a powerful predictor for worsening heart failure. This breakthrough device designation is a pivotal milestone in Noah Labs Vox's path to driving early access to life-saving interventions for patients with life-threatening diseases.— Oliver Piepenstock, CEO and co-founder, Noah Labs
Five trials, one central question
The credibility of any AI diagnostic rests on how many independent groups have tested it. Noah Labs Vox has been validated in more than five multicenter trials, in collaboration with Mayo Clinic, the University of California San Francisco, Charité Berlin, Maastricht University Medical Center, and Hospital Clínic de Barcelona.
The prospective PRE-DETECT-HF trial, running with EU funding across European university medical centers, is the one Noah Labs says will underpin EU approval. The company expects EU MDR certification by mid-2026, with an FDA trial scheduled to follow the breakthrough designation.
What the breakthrough designation actually changes
The designation does not clear the device. It expedites review of a future premarket approval, 510(k), or De Novo submission, and gives Noah Labs an interactive relationship with the agency during development. It is a strong signal, not a market authorization.
Oliver Piepenstock built the company with Marcus Hott and Dr. Leonhard Riehle in 2021, and it already sells a CE-marked platform, Noah Labs Ark, across European clinical sites. Vox is the diagnostic layer on top of that telemonitoring base, backed by $12 million in combined public grants and private equity, including European Innovation Council funding.
The field's first casualty
Vocal biomarkers are not a new idea, and the earlier wave did not all survive contact with the market. Kintsugi, a well-funded mental-health voice-biomarker startup that had published a depression-screening study in the Annals of Family Medicine, ceased commercial operations in February 2026 and open-sourced its technology after concluding that the regulatory and reimbursement path was too costly to sustain as a standalone business.
As of 2026, no vocal-biomarker software has received FDA clearance as a diagnostic device. The FDA's revised General Wellness guidance, issued in January 2026, drew a harder line between wellness positioning and claims that guide clinical care. A startup that blurs that line faces the same wall Kintsugi hit.
The counterweight is that the science is consolidating. In July 2026, the eVoiceNet and Bridge2AI-Voice consortia published the first international consensus framework defining what counts as a vocal biomarker, a vocabulary that regulators, engineers, and clinicians can actually share. Standards like this are what let a field move from scattered pilots to something insurers will price.
Where the money sits
GlobalData puts the patient-monitoring market at $21.4 billion in 2025, growing to about $25.9 billion by 2035. Cardiovascular monitoring is a concentrated market. GE HealthCare led it in 2025 with a 26.6% share, followed by Suzuken and Baxter. Abbott's implanted CardioMEMS line is the direct incumbent for remote heart-failure monitoring, and its next-generation CardioMEMS Hero cleared the FDA in March 2026, the same month as Noah Labs' designation.
EU MDR certification of Noah Labs Vox, expected mid-2026
The FDA trial following the breakthrough designation and its eventual 510(k) or De Novo filing
Whether a health system buys Vox as an overlay on existing remote-monitoring programs
How vocal-biomarker standards translate into reimbursement codes
What happens if voice becomes a standard cardiac vital sign?
Probability: 55%. The regulatory path is already de-risked by the breakthrough designation, and the clinical validation is multicenter and published.
✅ Arguments for
Five multicenter trials and a consortium-led standards framework reduce the evidence gap that killed earlier startups.
Confirmation criteria: EU MDR approval lands by mid-2026 and a US trial enrolls on schedule.
❌ Arguments against
Kintsugi's collapse shows the reimbursement wall is real, and the tightened wellness guidance raises the compliance bar.
Disconfirmation criteria: the EU or FDA trial misses on sensitivity, or the field stays stuck in pilot stage without reimbursement codes.
Development scenarios
🟢 Optimistic scenario (25%)
Implications: voice becomes a reimbursed screening layer, and the vocal-biomarker category validates its thesis across cardiac and cognitive conditions.
🟡 Base-case scenario (50%)
Implications: Noah Labs builds a profitable niche, and the category grows steadily without displacing incumbents like Abbott.
🔴 Pessimistic scenario (25%)
Implications: the voice-diagnostic thesis stalls, and capital retreats to more conventional monitoring hardware.
The investment read
For a principal evaluating this space, the distinction to hold is between a validated sensor and a validated algorithm. Abbott's CardioMEMS is the proven incumbent, with an FDA-cleared implant and an NHS recommendation. Noah Labs is the challenger betting that the cheapest, most scalable signal, the human voice, can catch the same deterioration without the surgery.
That bet is plausible enough for the FDA to fast-track it and for five university medical centers to stake their reputations on it. Whether it converts into a reimbursed business is the question the next two trials answer.