Hybrid Brain Machine interface

A brain machine interface (BMI) is a system that directly decodes the user's intention from brain signals to control a robot or an external device or stimulator. These systems have shown to be potential tools for the rehabilitation of motor-impaired patients after stroke. Brain signals can be measured with non-invasive methods such as electroncephalography (EEG) or with higher resolution techniques like intracortical neural recordings. On the other hand, electromyographic (EMG) activity from the muscles has been widely utilized to control prostheses, rehabilitation robots or electrical stimulators, as it provides a more direct and robust measurement of the user's motion intention than non-invasive brain signals. However, EMG also has some limitations and some stroke patients might not have residual EMG activity. As a result, hybrid BMIs (hBMIs) combining brain and muscle signals to control a rehabilitation or assistive device have been recently proposed as an improved and more robust technology than only-brain or only-muscle based interfaces. We have developed a novel hBMI based on EEG and EMG that aims at eliciting rehabilitation by acting both at the brain and muscle levels, which are both compromised after stroke. We have tested it in a preliminary study with stroke patients showing encouraging results (Sarasola-Sanz, 2017) and are currently working on its testing in a larger population as well as on more accurate invasive hBMI systems.

Schematic of the Hybrid Brain Machine interface

Neural correlates of motor rehabilitation and statistical models

Interhemispheric coherence could serve as an advanced marker of stroke impairment and recovery

Uncovering brain processes that cooccur with or predict clinical motor improvement is an important building block for the development of stroke rehabilitation schemes (for chronic and severely paralyzed patients in particular). Only a comprehensive model of the patients can guarantee success in a personalized and efficient stroke rehabilitation. Our research strives to advance the current state of the art by way of investigating the neural brain state, the psychological condition and the behavior of patients. We use advanced signal processing, data analysis and machine learning techniques to uncover neural correlates and markers of motor rehabilitation and develop statistical models reflecting the state of the patient. We evaluate these models in clinical trials.

Enhancing motor rehabilitation

Rich virtual environments could be used to elevate engagement

Gamification is a principle that introduces aspects of games into real-life environments. The areas in which it helps humans are manifold. Project and task management is only one example field that benefits from increased effectivity through gamification. Transforming rehabilitation into a game partly or fully could potentially elevate engagement and motivation of the patients and thus allow for intensification of the training and an increase in outcome. We are studying the potential of gamified rehabilitation in our group through clinical trials in collaboration with leading experts of the German video game industry, specialized on serious games.

Software development

Real-time 3D visualization of an exoskeleton for patient guidance

Most of the software we use for signal processing and data analysis is developed in-house to meet high demands. Technologies employed at our institute range from high-level analysis in Matlab, statistics in R and machine learning in Python to lower-level software development for laboratory environments in Go and C/C++. Testing our hypotheses in clinical trials calls for customized software implementation for proprioceptive Brain-Machine interfaces (e.g. an interface for a robotic exoskeleton), visualization of data (e.g. real-time 3D rendering of a model of an exoskeleton for guidance of the patient) and other laboratory software (e.g. a multi-perspective video recording system for quality control of the data).

Sleep and Motor Learning

Sleep plays a major role in motor learning and consolidation. Previous studies have shown a direct relationship between brain signatures during sleep (i.e., spindles or slow waves) and motor learning (or other types of learning). In this context, brain-machine interfaces, in which we “select” where the learning (neuronal changes) occur, is the ideal tool to understand how sleep and learning interact with one another. This is of great importance for stroke subjects in which the impaired sensorimotor system might not be capable of efficiently consolidating the newly encoded information. Interestingly, presenting task-related cues during certain phases of the sleep has been proven to help the consolidation of the previously learned task. This tool, commonly known as targeted memory reactivation (TMR), provides an interesting tool to further improve motor rehabilitation of stroke patients.

Closed Loop Gait Rehabilitation

This research line is focused on gait rehabilitation for patients with motor dysfunctions in lower limb. We aim to investigate and identify the neuroplasticity mechanisms occurring at the brain and spinal cord level which are responsible for motor recovery. Combining our broad knowledge in neuromodulation techniques and a wide variety of technologies for neurostimulation, the main objective is to propose a therapy to restore locomotor control in patients with neurological disorders.


This research line is focused on understanding the interaction between the nervous system and the electric and magnetic fields that are currenlty used to stimulate it such as Transcranial Magnetic Stimulation (TMS) and/or Functional Electrical Stimulation (FES). We use the Sim4life simulation platform to study the distribution of the electromagnetic field in the human body using accurate 3D human models and single neuron analytical models (Sim4life). Parameters such as frequency, amplitude, locations or dose are studied to improve and optimize currently used clinical procedures.

Calculation of the electric field distribution in the body using a TMS coil