Publication
Robust Online Movement Prediction from EEG Data for Post-Stroke Rehabilitation
Niklas Kueper; Sujin Kim; Mathias Trampler; Marc Tabie; Kartik Chari; Julian Fabricius; Elsa Andrea Kirchner
In: keine Angaben. Brain-computer interface research: a state-of-the-art summary 13. Springer Nature, 2026.
Abstract
Stroke is one of the leading causes of long-term disabilities and requires effective sensory motor
rehabilitation in the early post stroke period to provide rehabilitation therapy as required. While
the use of the non-invasive electroencephalogram (EEG) in Brain-Computer Interfaces (BCIs)
has been studied already for decades, approaches that enable reliable detection of movement
intentions from EEG are still missing or unsatisfactory. Therefore, we developed two approaches
to improve reliability and practical applicability in the detection of human movement intentions
from EEG signals to extend robot-driven post-stroke rehabilitation approaches. The first
approach was developed for the continuous detection of EEG-based movement intentions to
support post-stroke rehabilitation using the RECUPERA exoskeleton developed at our institute.
In this approach, two neural network models were applied for a robust online detection of
human movement intentions from EEG. The second approach was developed to avoid a
dedicated calibration session, which is usually required to train an EEG classifier during a
therapy session. In this approach, we used transfer learning to improve the practical
applicability of the BCI in real therapy sessions. Both approaches were combined and integrated
into a realistic virtual kitchen scenario to provide a comprehensive therapy approach with
gamification providing a contextual environment and enabling reliable interpretation of noninvasive EEG. Although the presented approach yields promising first results, the approach has
only been tested with healthy subjects so far.