Heart Failure
Conditions
Brief summary
VAPP-HF is a prospective, multi-center, observational study assessing whether daily voice recordings analyzed by a machine learning algorithm can detect changes in pulmonary arterial (PA) pressure in heart failure patients with implanted PA pressure sensors (e.g., CardioMEMS, Cordella). Patients across three sites in Germany and the United States provide daily voice recordings via a mobile app for 12 weeks while continuing standard PA pressure monitoring and heart failure care. Voice data is analyzed retrospectively after study completion; no clinical decisions are based on voice analysis during the study. The primary endpoint is the sensitivity and specificity of the AI-based voice analysis in detecting PA pressure changes at defined thresholds.
Detailed description
Implanted PA pressure sensors enable early detection of heart failure decompensation but are costly and invasive. Fluid retention in heart failure may affect the vocal apparatus, producing measurable voice changes that could serve as a non-invasive alternative for monitoring pulmonary congestion. Participants record daily voice samples consisting of sustained vowel sounds and a standardized reading passage via the Noah Labs mobile app. PA pressure readings are collected daily per standard care. Voice recordings and clinical data are analyzed retrospectively using classical machine learning and deep learning approaches. No additional clinical visits are required.
Interventions
OTHERDaily Voice Recording
Patients record daily voice samples (sustained vowels and a standardized reading passage) using the Noah Labs mobile app. PA pressure readings are collected daily per standard care using the implanted sensor. Voice recordings are analyzed retrospectively using machine learning algorithms after study completion.
Sponsors
Noah Labs
Study design
Observational model
COHORT
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* Age 18 years or older * Successful implantation of a PA pressure sensor and monitored by a participating study center * Willingness to record a short predefined text daily for 3 months using a smartphone or tablet * Ability to comfortably read aloud the study passage in English or German * Written informed consent obtained
Exclusion criteria
* Pregnant, breastfeeding, or unwilling to practice birth control during participation * Condition that in the opinion of the investigator would compromise patient safety or data quality * Pathological voice changes due to surgery or injury * Planned invasive cardiac procedures during the study period * COPD requiring home oxygen therapy * Chronic kidney disease requiring dialysis * Cognitive dysfunction limiting ability to perform daily voice recording * Inability to read English or German * Physical inability to use the recording device
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Sensitivity of AI Voice Analysis in Detecting PA Pressure Changes | 12 weeks | Sensitivity and specificity of the AI-based voice analysis algorithm in detecting pulmonary arterial pressure changes at pre-specified thresholds. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| orrelation Between Voice Predictions and Clinical Events | 12 weeks | Correlation between voice biomarker predictions and clinical outcomes including hospitalizations and diuretic adjustments. |
| Predictive Accuracy of Machine Learning Models | 12 weeks | Predictive accuracy of machine learning models for early detection of signs of heart failure decompensation, reported as area under the ROC curve. |
| Adherence to Daily Voice Recording | 12 weeks | Percentage of days with at least one transmitted voice recording over the 12-week study period. |
Countries
Germany, United States
Contacts
CONTACTLeonhard Riehle, MD
Outcome results
None listed