Blog
ACADEMICHealthcare + AI

Can AI Really Detect Heart Failure From Your Voice?

Noah Labs got FDA Breakthrough Device Designation for a voice-based heart failure detector. But does the science hold up? I ran it through APEX — here's what 35 sources say.

· 12 min read

The Claim

Noah Labs has developed a voice biomarker platform that claims to detect heart failure with 85% accuracy — and the FDA just gave it Breakthrough Device Designation. That's a significant regulatory signal. But clinical validation and regulatory designation are different things.

I tasked APEX with researching this from 35+ sources: peer-reviewed literature, FDA filings, clinical trial registries, and expert commentary. Here's what the data actually says.

What Voice Biomarkers Actually Measure

The science behind voice-based cardiac detection is rooted in cardiorespiratory coupling — the mechanical and neural interactions between the cardiovascular system and the respiratory system that subtly alter vocal cord tension, breath support, and phonation patterns.

When cardiac output decreases (as in heart failure), several measurable changes occur:

  • Reduced subglottal pressure from decreased respiratory muscle perfusion
  • Altered fundamental frequency (F0) variability
  • Changes in harmonic-to-noise ratio (HNR)
  • Subtle perturbations in jitter and shimmer measurements

These are real, measurable acoustic features. The question is whether they're specific enough to heart failure to be clinically useful.

What the Literature Actually Shows

APEX surfaced 14 peer-reviewed studies on voice biomarkers for cardiac conditions. The findings are more nuanced than the press releases suggest.

StudySample SizeAccuracyLimitation
Maor et al. (2020)108 patients82% AUCSingle-center, controlled environment
Solan et al. (2022)240 patients79% AUCExcluded patients with voice disorders
Noah Labs Phase II316 patients85% AUCNot yet peer-reviewed

The accuracy numbers look promising. But the limitations are significant: controlled recording environments, exclusion of confounding voice conditions (COPD, laryngitis, smoking history), and no external validation cohorts.

The FDA Designation: What It Actually Means

Breakthrough Device Designation does not mean the device works. It means the FDA has agreed to expedite its review because the condition is serious and the device may offer a substantial improvement over existing alternatives.

The designation accelerates the path to market — it doesn't validate the science. The device still needs to pass a De Novo or 510(k) pathway with clinical evidence.

The Clinical Perspective

As someone who has worked in cardiac monitoring, the baseline problem is critical to flag: voice biomarkers will need to be calibrated against each patient's baseline voice. A patient with a history of smoking, GERD, or even seasonal allergies will have different acoustic baselines than a healthy control.

The studies so far have largely used healthy controls as comparison groups — not patients with comorbidities. In a real clinical environment, the signal-to-noise ratio gets much worse.

Bottom Line

The science is real. The technology is promising. The current evidence is insufficient for clinical deployment without significant external validation.

What to watch: The Noah Labs Phase III trial results, expected Q3 2026. That's the dataset that will determine whether this is a genuine diagnostic tool or an expensive screening novelty.


This article was researched using APEX — an autonomous intelligence engine that cross-references 35+ sources, grades its own confidence, and flags what it doesn't know.

SoloBuilder Weekly

Join solo builders. Free, weekly, no spam.