53
Research Trials
20
Peer-reviewed publications
16
Clinical Conditions
This narrative review by Hyfe's R&D team makes the case that continuous cough monitoring (CCM), powered by acoustic AI, transforms cough from a subjective symptom into a quantifiable digital biomarker.
This ERS 2025 abstract from Hyfe's R&D team validates the wear-detection algorithm built into the Hyfe CoughMonitor smartwatch against participant-reported wear status across 418 person-hours of data in ten participants.
Authored by Hyfe's R&D team, this review synthesizes work presented at the European Respiratory Society Congress 2025 and argues that objective cough monitoring has crossed a practical threshold, moving from experimental technique to deployable clinical endpoint.
This study asked whether the core components of BCST could be embedded in a digital therapeutic and paired with continuous, objective cough monitoring inside the CoughPro app.
01.10.2024
Hyfe identifies and timestamps coughs thus providing continuous hourly cough counts. Hyfe uses the device's microphone and a two-layer AI system: (1) peak-detection, (2) cough classification, to detect and classify "cough-like" sounds on-device in a privacy preserving way. The retrospectively analyzed dataset was comprised of 97 participants, who monitored for 30 days, with >20 daily monitoring hours and a cough frequency >5 coughs per hour. The data, gathered between January and August 2023, included only cough timestamps and device usage times, with no additional user information.
We calculated daily cough frequencies by dividing total daily coughs by monitoring time and applied bootstrapping to hourly counts to establish 95% confidence intervals for each day. In assessing cough predictability, we calculated One Day's Predictability as the percentage of other days with cough frequencies within that day's 95% confidence interval. Overall Predictability, the mean of these percentages across 30 days, reflects the predictability of 24 hour monitoring. High values indicate a stable and predictable cough pattern, while low values suggest variability from day to day, and thus that 24 hour monitoring in inaccurate.
The mean (median) daily cough rates varied from 6.5 to 182 (6.2 to 160) coughs per hour, with standard deviations (interquartile ranges) varying from 0.99 to 124 (1.30 to 207) coughs per hour among all subjects. There was a positive association between cough rate and variability, as subjects with higher mean cough rates (OLS)have larger standard deviations. The accuracy of any given day for predicting all 30 days is the One Day Predictability for that day, defined as percentage of days when cough frequencies fall within that day’s 95% confidence interval. Overall Predictability was the mean of the 30 One Day Predictability percentages and ranged from 95% (best predictability) to 30% (least predictability).
Limitations: The clinical data was not available for most of the subjects. Although the cough detection algorithms have been extensively tested, their performance has not been validated for this use case.
Take-Home Points:
