A Power-Efficient Neuromorphic Digital Implementation of Neural–Glial Interactions

Throughout the last decades, neuromorphic circuits have incited interest of scientists, as they are potentially a powerful tool for treatment neurological diseases. To this end, it is essential to consider biological principles CNS and develop appropriate area- power-efficient circuits. Motivated by studies that outline indispensable role astrocytes in dynamic regulation synaptic transmission their active contribution neural information processing CNS, work we propose digital implementation neuron–astrocyte bidirectional interactions. In order describe neuronal dynamics astrocytes’ calcium dynamics, modified version original Izhikevich neuron model was combined with linear approximation Postnov functional neural–glial interaction model. For computation core, only three pipeline stages 10.10 fixed point representation were utilized. Regarding results obtained from FPGA comparisons other works, proposed circuit reported significant savings area requirements (from 22.53% up 164.20%) along considerable total power consumption 28.07% without sacrificing output accuracy. Finally, an RMSE analysis conducted, confirming particular produces more accurate compared previous studies.