Cardiac arrhythmia classification using wavelets and Hidden Markov Models - a comparative approach.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Cardiac arrhythmia classification is crucial for diagnosing heart conditions.
• Accurate feature extraction is essential for reliable arrhythmia detection.
• Conventional methods like linear segmentation have limitations in capturing complex signal variations.

Purpose of the Study:

• To comparatively evaluate standard linear segmentation and wavelet-based feature extraction for cardiac arrhythmia classification.
• To assess the performance of Hidden Markov Models in classifying different types of heartbeats.
• To identify the most effective feature extraction technique for improving diagnostic accuracy.

Main Methods:

• Utilized the MIT-BIH Arrhythmia Database for real-world data.
• Implemented Hidden Markov Models for beat classification.
• Applied standard linear segmentation and multi-scale wavelet transform for feature extraction.
• Classified normal (N), premature ventricular contraction (V), supra-ventricular arrhythmia (S), atrial fibrillation (AF), and atrial flutter (AFL) beats.

Main Results:

• Wavelet-based feature extraction demonstrated superior performance over standard linear segmentation.
• The multi-scale observation approach in wavelet transform effectively captured signal characteristics.
• Hidden Markov Models achieved robust classification with the proposed feature extraction methods.

Conclusions:

• Wavelet transform is a more effective feature extraction method for cardiac arrhythmia classification than linear segmentation.
• This approach enhances the potential for accurate and reliable diagnosis of various heart rhythm disorders.
• The findings support the integration of advanced signal processing techniques in clinical cardiology.