ECG SIGNAL CLASSIFICATION USING PTBXL DATASET using DL

The increasing prevalence of cardiovascular diseases has necessitated the development of automated diagnostic systems for early detection and accurate classification of heart conditions. Electrocardiography (ECG) is a widely used non-invasive technique for heart health monitoring. However, interpreting ECG signals manually can be time-consuming and error-prone. This project focuses on the classification of ECG signals using deep learning techniques, leveraging the publicly available PTB-XL dataset. The dataset consists of over 21,000 ECG recordings, along with detailed metadata about the patients. In this research, a multi-label classification approach is adopted, where ECG signals are categorized into five diagnostic classes: Normal (NORM), Myocardial Infarction (MI), ST/T Change (STTC), Conduction Disturbance (CD), and Hypertrophy (HYP). The project applies various deep learning models, including Convolutional Neural Networks (CNN), CNN + Long Short-Term Memory (LSTM), CNN + Variational Autoencoder (VAE), and CNN + Bidirectional LSTM with Attention Mechanism, to classify these signals. A combination of metadata and raw ECG data is used to create a comprehensive model for accurate prediction. The final model aims to provide reliable and efficient classifications that can assist healthcare providers in diagnosing cardiac conditions. Evaluation metrics such as binary accuracy, precision, and recall are used to assess the model’s performance. The results show that the CNN + Bidirectional LSTM model achieved the highest classification accuracy, making it a robust solution for ECG signal analysis. This research holds potential for automating cardiac diagnoses, reducing human error, and improving overall healthcare efficiency. 

Keywords: ECG classification, deep learning, PTB-XL dataset, CNN, LSTM, VAE, Bidirectional LSTM, cardiac health, Myocardial Infarction, multi-label classification.