Key result
The hybrid CNN-LSTM model using digital phonocardiogram spectrograms achieved 98.4% multiclass classification accuracy and 0.993 macro-average AUC across five cardiac conditions.
Why the study?
Cardiovascular diseases are a major global cause of death, creating an increasing demand for early and accurate diagnosis of cardiac diseases.
Does a hybrid CNN-LSTM deep learning framework using STFT spectrograms of digital stethoscope phonocardiogram signals accurately classify valvular heart diseases and normal heart sounds?
Observational (n=2,000)
No
Does a hybrid CNN-LSTM deep learning framework using STFT spectrograms of digital stethoscope phonocardiogram signals accurately classify valvular heart diseases and normal heart sounds?
Answers are grounded in this paper and cite their sources.
This study demonstrates that a hybrid CNN-LSTM deep learning model processing phonocardiogram spectrograms on edge devices can classify four valvular heart diseases and normal heart sounds with 98.4% accuracy.
Cardiovascular diseases (CVDs) are considered one of the major killers in the world and there is an increasing demand for early and accurate diagnosis of cardiac diseases. In this study, an Artificial Intelligence (AI) based smart cardiac health monitoring system for real time cardiovascular disease detection and classification via Phonocardiogram (PCG) signal is proposed. Heart sounds are collected with a digital stethoscope and processed with Short-Time Fourier Transform (STFT) techniques to create spectrograms, normalize and filter the noise. A hybrid deep learning model known as Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) architecture is designed to achieve fully automated feature extraction and multiclass disease classification. Presented with Explainable Artificial Intelligence (XAI) techniques to enhance transparency and clinical interpretability. The proposed framework had good classification accuracy and strong real-time performance and noise resistance. The implementation of edge AI with embedded devices allows for easy and low-cost deployment of healthcare solutions, which are ideal for telemedicine, rural healthcare, home monitoring, and smart healthcare applications.
No takes yet. Share an insight, caveat, or question.
Vijaybaskar et al. (2026) conducted an observational in Cardiovascular diseases (Aortic Stenosis, Mitral Regurgitation, Mitral Valve Prolapse, and Mitral Stenosis) (n=2,000). Hybrid CNN-LSTM deep learning framework with STFT spectrogram analysis of phonocardiogram signals vs. Traditional machine learning and deep learning models (SVM, Random Forest, ANN, CNN, MobileNet, ResNet) and clinical echocardiography confirmation was evaluated on Multiclass classification accuracy for cardiac diseases. The hybrid CNN-LSTM model using digital phonocardiogram spectrograms achieved 98.4% multiclass classification accuracy and 0.993 macro-average AUC across five cardiac conditions.
Synapse has enriched 10 related papers. Consider them for comparative context: