Machine Learning in Neural Engineering

Department Empirical Inference
Max Planck Institute for Intelligent Systems

Brain-Computer Interfacing in Amyotrophic Lateral Sclerosis

While most healthy subjects are nowadays capable of operating a BCI that enables basic communication, those that stand to benefit most from this technology, e.g., patients in late stages of amyotrophic lateral sclerosis (ALS), remain incapable of utilising a BCI. Furthermore, even subjects that are in principle capable of operating a BCI show a large variation in performance over the course of an experimental session. These factors impede a successful translation of BCI research into clinical and home use.

In order to realise reliable BCI-based communication, we first need to understand the neuro-physiological causes of across- as well as within-subject performance variations in patient populations. These insights can then be used to design new BCI systems that are robust to such variations and may enable BCI-based communication for severely paralysed patients in late stages of ALS. Accordingly, we use causal inference methods to investigate how brain regions interact in order to solve BCI-related tasks. We are particularly interested in the role of gamma-oscillations for information processing in the brain and their relevance for BCIs. This project has been awarded with the BCI Award 2011.

More details on this work can be found in the following selected publications:

Self-Regulation of Brain Rhythms in the Precuneus: A Novel BCI Paradigm for Patients with ALS (Journal of Neural Engineering, 2016) Two ALS patients reliably communicated with a BCI based on self-regulation of brain rhythms in the precuneus (resulting in the scalp topography shown to the left) over the course of 1.5 years.

A Brain-Computer Interface Based on Self-Regulation of Gamma-Oscillations in the Superior Parietal Cortex (Journal of Neural Engineering, 2014) Healthy subjects and one locked-in ALS patient learned to self-regulate the power of gamma-oscillations in superior parietal cortex by alternating between states of focused attention and relaxed wakefulness.

High Gamma-Power Predicts Performance in Sensorimotor-Rhythm Brain-Computer Interfaces (Journal of Neural Engineering, 2012) Differences in gamma-power between two fronto-parietal networks predict performance in a motor-imagery BCI on a trial-to-trial basis.

Causal Influence of Gamma-Oscillations on the Sensorimotor-Rhythm (NeuroImage, 2011) Distributed gamma-range oscillations have a causal influence on a subject's capability to modulate the sensorimotor-rhythm by means of motor-imagery.