Inter- and intra-patient ECG heartbeat classification for arrhythmia detection: A sequence to sequence deep learning...

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Signal Processing

Background:

• Electrocardiogram (ECG) signals are crucial for studying heart function and diagnosing arrhythmias.
• Current arrhythmia classification methods struggle with imbalanced datasets, limiting diagnostic performance.
• There is a need for robust automatic heartbeat classification methods.

Purpose of the Study:

• To develop an automatic heartbeat classification method using deep convolutional neural networks and sequence-to-sequence models.
• To address the limitations of existing methods in handling imbalanced datasets for arrhythmia detection.
• To achieve superior performance in classifying different heart conditions from ECG signals.

Main Methods:

• Utilized deep convolutional neural networks (CNNs) for feature extraction from ECG signals.
• Employed sequence-to-sequence models to enhance the classification accuracy of heartbeat patterns.
• Evaluated the proposed method on the MIT-BIH arrhythmia database using intra-patient and inter-patient paradigms, adhering to the AAMI EC57 standard.

Main Results:

• The method demonstrated state-of-the-art performance across both intra-patient and inter-patient evaluation schemes.
• Achieved high positive predictive values (e.g., 96.46% for S, 98.68% for F in intra-patient) and sensitivity (e.g., 100% for S, 97.40% for F in intra-patient).
• Further excelled in inter-patient analysis with excellent metrics for categories S (92.57% PPV, 88.94% sensitivity) and V (99.50% PPV, 99.94% sensitivity).

Conclusions:

• The proposed deep learning approach effectively classifies heartbeats and improves arrhythmia detection, particularly with imbalanced data.
• The method sets a new benchmark in the literature for ECG-based arrhythmia classification.
• This work offers a promising solution for more accurate and reliable cardiac condition diagnosis.