A Delay-Based Neuromorphic Processor for Arrhythmias Detection

Cardiovascular disease is the leading cause of global mortality, with 17.5 Million deaths per annum (World Health Authority, WHO). Innovative hardware based cardiac recording devices could help elevate this burden. Delay-based reservoir computing is a novel computational framework with only a single nonlinear node. This feature makes it a strong candidate for the hardware implementation of an analogue cognitive system. Such a system can be exploited to improve the energy efficiency of data processing in implantable bioelectronic devices. This paper presents a system modelling of this network that is capable of cognitively processing Electrocardiograph (ECG) signals from the MIT-BIH arrhythmia database. The proposed single-input single-output model receives an encoded ECG signal while the output amplitude pattern aids the diagnostic interpretation. The information processor is an analogue circuit with the dynamic properties of Mackey-Glass nonlinearity and fading memory. To validate this system and mimic real-time operation, the simulation is designed to detect ventricular ectopic beats, an ectopic heartbeat type, using a continuous ECG signal without any signal segmentation or feature extraction. After training, the model successfully locates ventricular ectopic beat with 87.51% sensitivity and 94.12% accuracy for the testing dataset from three patients.