ECGNet achieves exceptional performance in ECG signal classification, reaching approximately 96% accuracy on test data with a compact model of around 1300 parameters. This PyTorch-implemented model distinguishes between normal and abnormal heart rhythms using an efficient neural network architecture. With convolutional layers, batch normalization, dropout, and ELU activations, it demonstrates the potential of machine learning in enhancing cardiac diagnostics, all while maintaining a lean parameter count. This work underscores the balance between accuracy and model efficiency, leveraging the PTB Diagnostic ECG Database from Kaggle.
The dataset used in this project, specifically for training the ECGNet model, is part of the PTB Diagnostic ECG Database, available on Kaggle. This dataset is crucial for the project as it includes ECG recordings that have been labeled as normal or abnormal, providing a rich source of data for training and testing the model's ability to accurately classify ECG signals. The inclusion of this dataset supports the development of algorithms capable of distinguishing between different cardiac conditions, potentially contributing to advancements in diagnostic methodologies in healthcare.
For more detailed information about this dataset, please visit the Kaggle dataset page : https://www.kaggle.com/datasets/shayanfazeli/heartbeat