Neural Network Surrogates for Real-Time Optimization of High-Dimensional Neurostimulation

Neurostimulation is a medical and scientific technique that uses small electrical signals to communicate with the brain, spinal cord or nerves. It can help researchers understand how the nervous system controls movement, and it is already used in treatments for conditions like paralysis or Parkinson’s disease. However, one major challenge remains: to work well, stimulation must be carefully adjusted to each person. Today, this tuning is mostly done by trial-and-error, which can take a long time and often needs to be repeated.

This project aims to speed up and improve that process by developing artificial intelligence methods capable of automatically finding the best stimulation settings. Instead of testing thousands of possibilities manually, learning algorithms could quickly explore different options and adapt to each individual. The work focuses on fast optimization tools that can learn from data in real time, which could allow future neurostimulation therapies to become more efficient, personalized, and easier to use clinically.

If successful, this research could help expand access to neurotechnology, reduce the workload for physicians, and offer patients more responsive and reliable stimulation treatments. The goal is to support the next generation of intelligent neurostimulation systems that improve recovery and quality of life.

Faculty Supervisor:

Marco Bonizzato

Student:

Partner:

Scuola Superiore Sant'Anna

Discipline:

Life Sciences

Sector:

Artificial Intelligence

University:

Polytechnique Montréal

Program: