Accurate arrhythmia classification using auto-associative neural network.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Cardiology

Background:

• Cardiac arrhythmias affect approximately 1 in 18 Americans, contributing to rising coronary heart disease rates.
• Early diagnosis of arrhythmias via electrocardiogram (ECG) signals is crucial for reducing patient mortality.
• Current diagnostic methods may require complex feature extraction, impacting efficiency.

Purpose of the Study:

• To introduce and evaluate an Auto Associative Neural Network (AANN) as a novel, feature-extraction-free classification method for cardiac arrhythmias.
• To demonstrate the effectiveness of AANNs in representing arrhythmia classes through stored network weights, achieving high data compression.
• To compare the performance of the proposed AANN technique against established methods like Dynamic Time Warping (DTW) for arrhythmia classification.

Main Methods:

• Application of Auto Associative Neural Networks (AANNs) for direct classification of segmented ECG beats.
• Utilizing stored weights of trained AANNs as class representative models for compression.
• Testing the AANN technique on four distinct arrhythmia classes from the MIT/BIH Arrhythmia database.
• Comparative analysis against the Dynamic Time Warping (DTW) template matching technique.

Main Results:

• The proposed AANN technique achieved an average classification accuracy exceeding 97%.
• A relative compression gain of over 90% was observed compared to the training data size.
• The AANN method demonstrated competitive performance against the state-of-the-art DTW technique.

Conclusions:

• Auto Associative Neural Networks provide an effective and efficient approach for cardiac arrhythmia classification without requiring manual feature extraction.
• AANNs offer significant data compression capabilities, making them suitable for large-scale ECG analysis.
• This novel technique shows potential for improving early diagnosis and monitoring of arrhythmias, contributing to better cardiovascular health outcomes.