NEUROMORPHIC COMPUTING BASED ON STOCHASTIC SPIKING RESERVOIR FOR HEARTBEAT CLASSIFICATION

Abstract

Heart disease is the leading cause of mortality worldwide. The precise heartbeat classification usually requires a higher number of extracted features and heartbeats of the same class may also behave differently in patients. This will lead to computation overhead and challenges in hardware implementation due to the large number of nodes utilized in reservoir computing (RC) networks. In this work, a reservoir computing-based stochastic spiking neural network (SSNN) has been proposed for heartbeat rhythm classification, enabling a patient adaptable and more efficient hardware implementation with low computation overhead caused by minimum extracted features. Only a single feature is employed in template matching to achieve patient adaptability with minimal computation overhead. The single feature, QRS complexes, was extracted and fed into the neural reservoir with 20 neurons in a cyclic topology for arrhythmia similarity calculation and classification. 43 recordings of Electrocardiogram (ECG) signals that included both normal and arrhythmic beats from MIT-BIH arrhythmia database obtained from Physio-Net were used in this work. The proposed stochastic spiking reservoir achieves a sensitivity of 99.6% and an accuracy of 96.91%, signifying that the system is accurate and efficient in classifying normal and abnormal arrhythmias.