FPGA implementation of a Spiking Neural Network for Real-Time Action Potential and Burst Detection

Action potentials are the incompressible building blocks of the language of neurons. As a result, their detection in real-time is crucial for the understanding of neural populations in electrophysiology experiments. Higher-level features such as bursts provide as well interesting insights on neural computation. Because of the importance of this problem, a wide range of spike detectors have been developed for the analysis of data from probes such as Micro-Electrode Arrays (MEA). The ability to perform this task online and in real-time further opens potential application like closed-loop bio-hybrid experiments, electroceutics and Brain-Machine-Interfaces (BMIs) could follow this milestone. However, neuroprosthetics and BMIs demand drastic technological constraints in terms of portability, energy consumption and accuracy. We present a FPGA implementation of a Spiking Neural Network (SNN) for real-time online detection of action potentials and bursts as a basis for implementing further functionalities. It shows better result compared to usual methods and its compact architecture takes small amounts of logic, translating into limited power consumption and high level of portability.