MySteth AI Tool Detects Heart Murmurs at Home Using Smartphone and Stethoscope
New AI system achieves 92% accuracy in detecting heart murmurs using just a smartphone or digital stethoscope for home cardiac screening.
Summary
Researchers developed MySteth, an AI-powered tool that uses smartphones or digital stethoscopes to detect heart murmurs at home. The system achieved 92% accuracy in distinguishing normal heartbeats from murmurs and 91% accuracy in classifying murmur types. Using deep learning and widely available technology, MySteth could enable early detection of cardiac conditions in underserved areas where specialized equipment isn't accessible, potentially improving cardiovascular health outcomes through convenient home monitoring.
Detailed Summary
Cardiovascular disease kills 17.9 million people annually, making early detection crucial for prevention. However, many people lack access to regular cardiac evaluations, particularly in rural and underserved areas. Researchers have now developed MySteth, an innovative AI-powered diagnostic tool that enables home-based heart monitoring using just a smartphone or digital stethoscope.
The research team trained a hybrid CNN-LSTM deep learning model on 10,000 heart sound recordings, all meticulously annotated by specialized cardiologists. The system processes audio recordings to distinguish between normal heart sounds and various types of murmurs - abnormal sounds that can indicate valve dysfunction or other cardiac problems. MySteth goes beyond simple detection, classifying murmurs into systolic and diastolic categories, and further subdividing systolic murmurs into ejection systolic murmurs (ESM) and pansystolic murmurs (PSM).
The results demonstrate impressive clinical accuracy: 92% for detecting murmurs versus normal heartbeats, 91% for classifying murmur timing, and 90% for distinguishing between ESM and PSM types. This level of precision rivals traditional clinical assessment methods while requiring only equipment most people already own or can easily access.
The implications for global health are significant. MySteth could enable early detection of conditions like mitral regurgitation, aortic stenosis, and other valvular diseases before they progress to require invasive interventions like coronary artery bypass surgery. The tool's accessibility makes it particularly valuable for remote monitoring and screening in areas with limited healthcare infrastructure, potentially reducing the global burden of undiagnosed cardiovascular disease.
Key Findings
- MySteth achieved 92% accuracy in detecting heart murmurs using smartphone recordings
- System classified murmur types with 91% accuracy for systolic vs diastolic categories
- Tool distinguished between ejection systolic and pansystolic murmurs with 90% precision
- Deep learning model processed 10,000 cardiologist-annotated heart sound recordings
- Technology requires only smartphone or digital stethoscope for home cardiac screening
Methodology
Researchers used a hybrid CNN-LSTM neural network trained on 10,000 heart sound recordings from smartphones and digital stethoscopes. All audio samples were annotated by specialized cardiologists and processed using Linear Predictive Coding for feature extraction.
Study Limitations
The study used a relatively small initial dataset that required synthetic augmentation. Clinical validation in diverse populations and comparison with traditional diagnostic methods would strengthen the evidence base for real-world implementation.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.