Automatic arrhythmia detection based on time and time-frequency analysis of heart rate variability

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Arrhythmia detection is crucial for cardiovascular health management.
• Current methods may require complex feature extraction or extensive data.
• Developing efficient, automated systems for arrhythmia detection is an ongoing need.

Purpose of the Study:

• To develop an automatic arrhythmia detection system utilizing solely heart rate features.
• To evaluate the efficacy of time and time-frequency domain analyses for arrhythmia classification.
• To assess the performance of neural networks trained on extracted cardiac signal features.

Main Methods:

• Extraction of RR interval duration signals from ECG recordings.
• Segmentation of RR interval signals and analysis using time and time-frequency domain features.
• Training of neural networks using combinations of extracted time and time-frequency features, including Short Time Fourier Transform and various Time-Frequency Distributions (TFD).

Main Results:

• The system demonstrated satisfactory sensitivity and specificity on the MIT-BIH arrhythmia database.
• Time domain analysis yielded 87.5% sensitivity and 89.5% specificity.
• Time-frequency domain analysis achieved higher performance with 90% sensitivity and 93% specificity.

Conclusions:

• An automatic arrhythmia detection system based solely on heart rate features is feasible and effective.
• Time-frequency analysis offers improved performance over time domain analysis for this application.
• The developed system shows promise for accurate and automated arrhythmia detection using ECG data.