Robust Time-Based ECG Slope Detection System Utilizing Izhikevich Neurons Eliminating Preprocessing: Resilience to Baseline Wander and Noise

Arrhythmia characterized by irregular heartbeats poses significant health risks and necessitates early diagnosis and treatment. While existing methods for arrhythmia detection are frequency based or ANN based techniques and it typically involve several stages of preprocessing and filtering of ECG signals. This paper presents a novel time-domain-based approach that operates directly on raw ECG data without the need for preprocessing enhancing the accuracy and reliability of arrhythmia detection. Izhikevich neuron model allows encoding raw ECG signals into spikes based on amplitude, enabling the extraction of timing features by detecting slope changes. This approach effectively identifies distinct slope variations corresponding to ECG waves (P, QRS, T) despite the presence of noise and baseline wander. The choice of the Izhikevich neuron model is motivated by its balance between biological realism and computational efficiency. Experimental validation using waveforms from the MIT-BIH arrhythmia database demonstrates the robustness of the proposed system, which has been implemented on a Zynq-7000 FPGA board.