[2023] GPU acceleration of neural-interface processing for closed-loop brain stimulation

One popular approach to understanding brain function is by studying electrophysiological recordings from animals. For high-resolution recordings (i.e., the ones using multi-electrode arrays or MEAs), the volume of the recorded neural data becomes significant, which makes it difficult to analyze and consequently construct a suitable stimulus back to the brain. Such a closed-loop design would help answer very critical questions; e.g., how do body movements in different time scales affect the neural signals generated in the prefrontal cortex? Preliminary experiments indicate such signals to compress or stretch in time to accommodate for reciprocal body movements.  Understanding the underlying mechanisms of this phenomenon are crucial for, ultimately, tackling sensorimotor diseases in humans and, even, facilitating next-generation brain-machine interfaces (BMIs).

Assignment

This thesis comprises three stages: 1) Experiment with the open-ephys 2 platform (https://open-ephys.org) and its potential for building fast neural loops. 2) Extend the system by adding a powerful GPU in the mix, which will be used to speed up calibration of animal-recording data for optimal closed-loop neurostimulation. 3) (Optional) Devise simplified artificial-neural-network (ANN) models to approximate realistic Spiking brain models, and then port them to the same GPU to mount true closed-loop experiments.

Requirements

MSc student in Computer Engineering, Embedded Systems or Biomedical Engineering with a profile in signal processing and machine/deep learning, C/C++ programming, algorithms, parallel programming, Linux. Optionally, being familiar with FPGA design.

Contact

Last modified: 2022-10-04