Automatic Arrhythmia Diagnosis System Based on CNN-LSTM

Electrocardiograms (ECG) are essential tools for diagnosing cardiovascular diseases, yet traditional ECG identification relies on the experiential interpretation by doctors, which is time-consuming and subject to subjective influence. In light of the limitations of ECG, this study introduces an automatic ECG signal analysis system based on Convolutional Neural Networks (CNN) and Long Short-Term Memory networks (LSTM). This system is capable of learning from and recognizing complex patterns in vast ECG datasets. Initially, the study utilizes four public ECG databases, and the data are denoised, sliced, resampling and combined. Subsequently, a CNN-LSTM model is applied to perform seven-category classification of the ECG signals. In the end, the CNN-LSTM model showed a 98.20% recall rate, 98.30% accuracy, and 98.78% accuracy. Additionally, the fully trained model has been successfully deployed on a Raspberry Pi, enabling online applications and real-time diagnostic capabilities. Simultaneously, this study also developed a wearable ECG monitoring vest based on the AD8232. This vest is capable of real-time ECG signal acquisition and transmits the data to the Raspberry Pi via Bluetooth technology, facilitating immediate detection of cardiac anomalies.