Design and Optimization of ECG Modeling for Generating Different Cardiac Dysrhythmias.

Area of Science:

• Biomedical Engineering
• Computational Biology
• Signal Processing

Background:

• Electrocardiograms (ECGs) are crucial for diagnosing cardiovascular diseases.
• ECG morphology varies significantly with cardiac conditions, necessitating accurate modeling.
• Existing modeling methods face challenges in accuracy and parameter localization.

Purpose of the Study:

• To develop a simplified ECG model using a minimal parameter set for accurate representation of various cardiac dysrhythmias.
• To introduce and evaluate two novel hybrid optimization techniques, ApproxiGlo and ApproxiMul, for fitting the ECG model parameters.
• To explore potential applications of the developed ECG model, including generation and data compression.

Main Methods:

• A mathematical ECG model based on the sum of two Gaussian functions was proposed.
• Two hybrid optimization methods, ApproxiGlo (approximation + global search) and ApproxiMul (approximation + multi-start search), were developed.
• The model and optimization methods were applied to real ECG data from various cardiac dysrhythmias and evaluated in time, frequency, and time-frequency domains.

Main Results:

• The proposed ECG model achieved high correlation coefficients (>0.98) when fitting diverse ECG beats representing different cardiac dysrhythmias.
• ApproxiGlo and ApproxiMul demonstrated significant improvements in root mean square error (RMSE) compared to nonlinear fitting, being 3.32 and 7.88 times better, respectively.
• The ApproxiMul optimization method outperformed ApproxiGlo across multiple performance metrics. The model achieved a 20:1 compression ratio for ECG data.

Conclusions:

• The simplified Gaussian-based ECG model, optimized by hybrid methods, accurately represents ECG morphology in various cardiac dysrhythmias.
• The developed optimization techniques offer superior performance over traditional methods for fitting Gaussian sums.
• The model has practical applications in syntactic ECG generation and efficient lossy data compression, with potential use in other engineering fields.