Title: Machine Learning for Robust and Scalable Neurotherapies (project of Pedro Abranches)
Electrical stimulation of the spinal cord mediates neurological recovery in humans with spinal cord injury. However, this therapy relies on a team of human experts that configures the temporal and spatial features of stimulation protocols through lengthy experimental sessions. Here, we propose to use methods in applied mathematics and machine learning with advanced neurophysiological biomarkers to develop algorithms that automatically configure these stimulations.
The goal is to develop fast and automated procedures that can deal with the combinatorial and high dimensional problems that EES promotes. This would enable us to deploy EES therapies in a worldwide fashion.
More specifically we use Bayesian optimization (BO) techniques to deal with the first part of the problem, which is finding single EES protocols. However, to promote a walking pattern, we need to combine several of these protocols in a temporal fashion. This creates a combinatorial and high dimensional problem. To address those we are exploring methods based on graph theory.
We have already set up an automatic framework to do experiments with human subjects and already started our experiments. This framework will now be used as a test bed for next improvements, which will hopefully aid patients with spinal cord injury.
More information about the SV iPHD program here