Abstract
Accurate detection of arrhythmias from electrocardiogram (ECG) signals is crucial for timely diagnosis and effective management of cardiovascular diseases. This paper introduces a novel bio-inspired approach for ECG arrhythmia detection, leveraging Spiking Neural Networks (SNNs) inspired by biological neural mechanisms. The proposed methodology employs a structured pipeline, beginning with signal preprocessing involving normalization and filtering. Continuous ECG signals are then transformed into spike trains using rate coding. The core of the approach utilizes leaky integrate-and-fire (LIF) neurons in combination with spike timing-dependent plasticity (STDP), modeling synaptic plasticity observed in biological neurons. The network dynamically updates synaptic weights based on the timing of input and output spikes, enabling it to learn complex temporal patterns from encoded ECG data. The SNN model was trained and evaluated using a comprehensive 12-lead ECG dataset aimed at classifying various arrhythmic conditions. The developed SNN achieved a high overall accuracy of 94.4% in arrhythmia detection tasks. Accuracy, sensitivity, and F1-scores exceeded 0.88 across all arrhythmia classes. Notably, the model demonstrated exceptional performance in identifying left bundle branch block (LBBB) and right bundle branch block (RBBB), attaining F1-scores of 1.00 and 0.99, respectively. The bio-inspired SNN approach effectively captures temporal dynamics critical for accurate arrhythmia classification. With high accuracy and reliability demonstrated across various arrhythmic conditions, particularly in distinguishing LBBB and RBBB, this model holds significant potential for enhancing automated ECG interpretation and supporting clinical decision-making.