by Lydia Timm, Ph.D.
Scientific Consultant (Brain Products)
On Saturday, October 8, 2016 something special will happen in Zürich, Switzerland. Scientists, developers and company representatives from all around the world will come together to see how their technological developments impact individuals with physical disabilities. “Cybathlon” can be seen as the new bionic Paralympics where people with severe disabilities can compete, helped with powered assistive devices. The competition comprises six different disciplines: a brain-computer interface (BCI) race, a functional electrical stimulation bike race, a powered arm prosthesis race, a powered leg prosthesis race, a powered exoskeleton race and a powered wheelchair race.
After Brain Products announced its support for the Cybathlon 2016 (in October 2014) we had many impressive and inspiring requests concerning loan equipment for the different groups.
Team Imperial (Imperial College London, UK) is involved in 4 of the races, including the BCI race. In this race, the pilots compete in a video game-based race controlled by their minds.
Team Imperial is using an EEG acquisition system (actiCHamp) to acquire brain activity from the pilot as he performs motor imageries (imagination of motor movements) corresponding to the command he wants to perform: e.g. thinking about clenching the fist to jump. These signals are decoded in real-time into one of the four different possible commands with a state-of-the art machine learning algorithm. As a screen shows the decoded commands, the pilot can learn from the feedback and improve his performance.
In order to decode these signals, Team Imperial will use a machine learning algorithm called “convolutional neural network” (CNN). This algorithm has proven itself very efficient in the field of image recognition and can be adapted to work with time-series. Based on the way the visual cortex recognizes objects, the CNN learns patterns in the EEG signals when provided with enough training examples. For each one of the possible commands, the network will be fed with thousands of examples of what EEG signals corresponding to this particular command may look like. However thousands of examples may not be enough and signal features could be required in order to reduce the dimensionality of the problem. This said, the performance of the whole system does not depend only on the accuracy of the algorithm but also on the training of the subject. When playing the game, the subject will receive feedback of his actions from the screen. This closed-loop system allows him to act indirectly on his brain signals as they get more and more correlated with the intended commands. → more about this team
The Athena-Minerva Team from the Technical University of Darmstadt (Germany) is also competing in the BCI race discipline. The team developed a BCI system enabling mind-based control of a video game for a disabled person who they call “the pilot”. The team’s BCI system consists of state-of-the-art artifact reduction methods and is capable of semi-automatic evaluation of different conditions to build personalized paradigms. The best performing paradigms are further improved by using adaptive neurofeedback methods.
Last but not least researchers from Pázmány Péter Catholic University (Faculty of Information Technology and Bionics, Electrophysiology Lab) in Hungary are planning to use a “hybrid” modality pattern for BCI feature extraction. They aim to use motor imaginary and complex function at the same time, and utilize complex algorithms based on wavelets and inverse ICA. Currently they are still evaluating their analysis methods given that some of their experiments showed that adaptive hysteresis based filtering increased response time. They are working with two tetraplegic (they are already accepted by the Cybathlon organizers) and two paraplegic patients who will compete in the BCI race. → more about this team
Brain Products wishes all competitors a successful tournament and the best of luck! We will be there to support and cheer for the competing teams with our CEO Alexander Svojanowski and other colleagues.