An IoT ECG Monitoring Framework for Arrhythmia Detection Using CNN–BiLSTM
الكلمات المفتاحية:
Electrocardiogram (ECG)، Arrhythmia Detection، Hybrid Deep Learning، CNN–BiLSTM Models، IoT-Based Monitoring، MIT-BIH Datasetالملخص
Cardiovascular disease is the leading cause of death worldwide, as many of these deaths are associated with sudden and unexpected heart rhythm disorders. This study presents an IoT-based integrated framework for ECG, based on a hybrid deep learning model that combines CNN and a Bidirectional Long Short-Term Memory (BiLSTM) to classify several types of arrhythmias, with the MIT-BIH Arrhythmia database to verify the performance of the model. Real-time ECG signals were captured with the AD8232 sensor and transmitted via the ESP32 microcontroller into a MATLAB-based processing and classification environment. The classification performance was evaluated under realistic conditions including the imbalance of categories, as the system achieved a total accuracy of 93.22%. In addition, the model was evaluated on the independent ECG5000 database, with a resolution of 90.10%, confirming the ability to generalize the model through different databases. A dedicated application for doctors and integrated with Firebase Cloud Messaging (FCM) has also been developed to send instant notifications when arrhythmia is detected.
