Deep Learning Framework for ECG and EEG Signal Classification in Smart Healthcare Systems

Abstract

Biomedical signal analysis is a key component of current healthcare systems for the diagnosis of cardiovascular and neurological diseases through Electrocardiogram (ECG) and Electroencephalogram (EEG) signals. This study aims to build an automatic deep learning-based classification system for ECG and EEG signals to facilitate the early diagnosis of cardiovascular and neurological disorders and to enable real-time monitoring in health care. Our approach employs two publicly accessible biomedical datasets, the MIT-BIH Arrhythmia dataset for ECG signals and the Epileptic Seizure Recognition dataset for EEG signals. Data preprocessing includes data cleaning, normalization, reshaping, exploratory data analysis, and data splitting into training, validation, and testing sets. We have developed multiple deep learning models such as Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) and Hybrid CNN-LSTM for biomedical signal classification. Evaluation metrics of accuracy, precision, recall and F1-score were used to assess model performance. The comparison of the results demonstrate that the CNN model has the highest accuracy of 0.98, Hybrid CNN-LSTM has 0.9767 accuracy, LSTM has 0.9751 accuracy and the RNN has 0.85 accuracy. The outcomes indicate that deep learning models can classify ECG and EEG signals and can be used for real-time embedded systems in health monitoring.