Neural Decoding of Walking in Stroke
MetadataShow full item record
Stroke is a leading cause of long-term disability in the U.S. and does not have an effective cure yet. Most post-stroke motor rehabilitation programs rely on physical therapy, which focuses on manipulating affected limbs. However, such programs only indirectly engage the cause of the symptoms: the damaged brain areas. Brain-computer interfaces (BCI) are systems that use brain activity to control external devices, allowing direct monitoring and involvement of patients' brains. To explore this novel stroke rehabilitation paradigm, three subjects with chronic stroke were recruited. They walked on a treadmill while controlling a virtual humanoid avatar on a TV through their brain signals. This real-time control was implemented based on a Kalman filter which predicted the subject's joint angles using their electroencephalogram (EEG) in the delta band (0.1--3 Hz). The performance was measured by Pearson's correlation coefficient r between the actual and predicted joint angles. This closed-loop BCI training paradigm was repeated for 16 trials during 8 days for each subject. Results show that promising performance was achieved (r=0.60, 0.59, 0.30 for the three subjects). Significant cortical adaptation was found in the delta and theta band EEG across trials. The delta band EEG, which drove the BCI decoder, was significantly associated with the decoder performance. Various motor learning effects were observed, including improved spatial and temporal gait consistency, and bilateral symmetry. The present study is the first BCI protocol designed for stroke rehabilitation in the context of walking. The promising BCI performance was backed up by significant correlations in EEG brain features, and more importantly, by the subjects' motor function improvement. The reported training paradigm may elicit several key adaptations in the EEG and motor skills of the chronic stroke survivors, which could lead to a new generation of stroke rehabilitation programs.