Classification of Electrocardiography Hybrid Convolutional Neural Network-Long Short Term Memory with Fully Connected...

Area of Science:

• Cardiology
• Artificial Intelligence
• Signal Processing

Background:

• Electrocardiography (ECG) is crucial for diagnosing heart rhythm conditions.
• Current methods for ECG analysis can be subjective and time-consuming.
• Automated analysis of ECG signals is needed for efficient and objective interpretation.

Purpose of the Study:

• To develop and evaluate a hybrid deep learning model for automated ECG analysis.
• To computerize the detection of normal and anomalous ECG patterns.
• To assess the performance of combined CNN, LSTM, and DNN architectures for ECG interpretation.

Main Methods:

• Utilized encoder-decoder techniques and loss distribution for anomaly prediction.
• Integrated Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Deep Neural Networks (DNNs) into a single architecture.
• Trained and validated the model using ECG data from the MIT-BIH arrhythmia database.

Main Results:

• The proposed hybrid model effectively describes ECG series and identifies abnormalities.
• Achieved superior performance in detecting both short-term and long-term abnormalities compared to existing methods.
• Demonstrated robustness in handling unbalanced datasets for ECG beat detection.

Conclusions:

• The integrated CNN, LSTM, and DNN approach offers a powerful tool for ECG interpretation.
• This technique can aid cardiologists in performing reliable and impartial ECG analysis, particularly in telemedicine.
• The model's ability to handle data imbalance and signal accuracy issues enhances its clinical utility.