Detection of Aortic Stenosis With Smartphone Auscultation Using Machine Learning (HEARTBEAT-Pilot)

Overview

Severe aortic stenosis, a common heart valve issue, is usually treated surgically or through intervention. Diagnosis typically occurs after symptoms appear, but research suggests already treating asymptomatic cases may help patients live longer. Current diagnostics using echocardiography are detailed but time-consuming, prompting the exploration of a smartphone application using built-in microphones and machine learning for quicker and more accessible screening.

Severe aortic stenoses usually is treated either surgically or interventionally, making it the most frequently treated among heart valve diseases. Typically, severe aortic stenosis is diagnosed only after the onset of the first symptoms. However, initial studies suggest that treating asymptomatic aortic stenoses could also extend the lifespan of affected individuals. Therefore, a widely applicable and cost-effective diagnostic method would be desirable for screening. The current gold standard for diagnosing aortic stenosis is echocardiography. It allows for detailed measurement and evaluation, assisting in detection and diagnostic assessment. However, it is time-consuming and therefore not readily applicable to a larger population. Alternatively, auscultation as an acoustic method is suitable, where typical noise changes due to turbulence in blood flow can be detected using a stethoscope. Since stethoscopes are only conditionally accessible for self-use, both in terms of availability and usability, this study aims to investigate whether a mobile application based on artificial intelligence for common smartphones using built-in microphones can also be diagnostically used. For this purpose, microphone recordings at the typical five auscultation points of 50 patients with severe aortic stenosis and 50 patients without any relevant heart valve disease are recorded. A digital stethoscope (3M Deutschland GmbH, Germany) and echocardiography findings serve as references. Based on the data, a classification model will be developed in a first step, which can detect severe aortic stenoses in smartphone recordings using machine learning.

Conditions

Interventions

Eligibility

Locations (1)

Outcomes