Key result
Memory-efficient classifiers for PCG signals achieve ~99% accuracy detecting cardiovascular disease on wearable devices.
Why the study?
Conventional diagnostic tools for cardiovascular diseases typically require expensive instrumentation and specialized staff, making inexpensive and non-invasive alternatives like phonocardiograms desirable.
Can machine learning classifiers accurately detect heart failure and other cardiac conditions from PCG signals while remaining memory-efficient for edge computing?
Population
10104 and 13136 5-s PCG signal frames from the Physionet 2016/CinC database
Comparison
Several ML/DL models including SVMs, k-NNs, and NNs for binary and multiclass classification
Design
Machine learning model development and validation study
Authors
Loading...
May support on-device PCG screening; leaves open prospective clinical validation before practice adoption.
Can machine learning classifiers accurately detect heart failure and other cardiac conditions from PCG signals while remaining memory-efficient for edge computing?
Memory-efficient neural networks can accurately classify normal and pathological phonocardiogram signals, enabling potential deployment on resource-limited wearable devices for heart failure detection.
Spongano et al. (2024) studied Cardiovascular diseases (heart failure, mitral valve prolapse, coronary artery disease). Machine learning classifiers (NNs, k-NNs, SVMs) for PCG signals was evaluated on Classification accuracy and F1-score. Memory-efficient neural networks and k-NN classifiers for PCG signals achieved up to 96.0% and 98.7% accuracy, respectively, for detecting cardiovascular diseases on wearable devices.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: