An Intelligent Framework for ECG-Based Cardiac Arrhythmia Classification Using Entropy-Guided Genetic Algorithm and Multi-Layer Perceptron Neural Network

Electrocardiogram (ECG) signal classification plays a vital role in the early detection and diagnosis of cardiac arrhythmias. In this study, we propose an efficient and lightweight classification pipeline that integrates wavelet-based preprocessing, entropy-guided feature selection using a Genetic Algorithm (GA), and a compact multilayer perceptron (MLP) for final classification. The ECG signals are first denoised using wavelet transform to remove baseline wander and high-frequency noise, after which R-peaks are detected and fixed-length segments are extracted around each peak. A total of ۲۸ handcrafted features-spanning time-domain, frequency-domain, wavelet, and higher-order statistical descriptors-are initially extracted from each heartbeat. The GA then selects an optimal subset of ۲۲ features to reduce dimensionality and enhance classification accuracy. The final MLP classifier is trained using these selected features to categorize heartbeats into five arrhythmia classes according to the AAMI EC۵۷ standard. The model was trained and evaluated on the MIT-BIH Arrhythmia Database using a patient-independent data split. Experimental results show that the proposed method achieves an average classification accuracy of ۹۹.۰۹%, with ۹۹.۱۲% sensitivity and ۹۹.۴۳% specificity across ۱۵ independent runs. The results demonstrate the model's robustness, computational efficiency, and suitability for real-time cardiac monitoring systems and wearable devices.

کلیدواژه ها:

ECG ، cardiac arrhythmia ، feature selection ، genetic algorithm ، multilayer perceptron