A Flexible FPGA Implementation of Morris-Lecar Neuron for Reproducing Different Neuronal Behaviors

Abstract

The Morris-Lecar (ML) neuronal model is one of the most popular biophysical models for studying the biological behaviors of single neurons. It has been implemented on hardware in different approaches. Field Programmable Gate Arrays (FPGA) technology has been recently used to implement different neuron models, which have a significant impact on new designs in the field of neuromorphic engineering. This interest is due to the FPGA's ability to implement massive and parallel architectures with high flexibility and accuracy. To date, different analog and digital designs have been proposed to implement the ML model on electronic circuits. In this work, we developed a real-time tunable architecture of the ML neuron on an FPGA. Using a noisy signal to stimulate the neuron, we tested the performance of the implemented ML model on both hardware and software models. We verified the FPGA model's capability to adjust various parameters in real-time. We used Xilinx ZCU102 FPGA boards to implement the real-time ML neuron. We discretized the ML model and updated FPGA model parameters for different ionic currents to reproduce neuronal activities generated by MATLAB. The hardware implementation demonstrated that the FPGA-based model reproduced the exact behavior implemented by MATLAB for different parameters and noisy injected current. Our design enabled adaptive alterations of the FPGA model neuron from Integrator mode to Differentiator mode in real-time. The Spike Triggered Average (STA) for several behaviors has been calculated.