Hybrid Excitable Spiking Neural Network for Cardiovascular Disease Prediction

Abstract

The functional interactions between the brain and heart through analysis of electrical activity (EEG and ECG) provide effective bio-markers for cognition, emotions, and morbidity. Conventional computational models designed for the brain and heart can be modified to understand the bidirectional interactions as the brain-heart interface for further clarification and treatment of morbidity. In this work, hybrid excitable cell model of neuron and cardiac myocyte is proposed and simulated with adrenergic features. The accuracy of proposed model observed 86% in terms of coincidence factor and 83% in terms of firing rate coefficient when compared with state-of-the-art computational models. The proposed model extended to the network of cardiac myocytes and electrical activity is recorded. The major findings of the study include i. Cardiac action potential heterogeneity is a significant factor in heart rate variability. ii. Spiking neural networks can accurately predict heart rate variability in real time from electrophysiological data of the cardiac network. iii. The electrical activity of the heart can be monitored and controlled by processing electrophysiological data of cardiac myocytes with spiking neural networks coupled with ion channels as voltage regulators to reduce the risk of cardiac morbidity and mortality. This work provides the initial phase of the brain-heart interface as a tool for the diagnosis of cardiac morbidity in real-time. The recent advancements in nano and bioelectronics will make it possible to deploy a brain-heart interface as a nano-chip to monitor and control the electrophysiological abnormality of the brain and heart by integrating nano-regulators with ion channels for stimulation.

Keywords

• Artificial intelligence
• feature extraction
• Izhikevich model
• neural networks
• neurocomputational features