Online co-adaptive brain-computer interfacing: Preliminary results in individuals with spinal cord injury

Abstract

Co-adaptive training paradigms for event-related desynchronization (ERD) based brain-computer interfaces (BCI) have proven effective for system setup and training of healthy users. However, there is little evidence as to whether co-adaptive ERD based BCI training paradigms could also benefit severely disabled users, including persons with spinal cord injury (SCI). Here, we present a preliminary study involving individuals with SCI at cervical level. In a cue-paced paradigm, our co-adaptive BCI analyzes the electroencephalogram from three bipolar derivations (C3, Cz, and C4), while the user alternately performs right hand movement imagery (MI), left hand MI and relax with eyes open. After less than five minutes of data collection, the BCI auto-calibrates and provides feedback for the MI task that can be classified better against relax with eyes open. The BCI then regularly recalibrates the underlying classifier model. In every calibration step, the system performs rigorous outlier rejection, selects the one out of six predefined logarithmic bandpower features (9 to 14 and 16 to 26Hz for the bipolars at C3, Cz and C4) that shows highest discriminability, and trains a linear discriminant analysis classifier. In under 30 min of training, all six tetraplegic users reached better than chance (p=0.01) online ERD based BCI control at an overall mean accuracy of 69.5 ± 6.4 %. These positive findings encourage us to evaluate the efficacy of adaptive BCI systems in users who have functional disability as a result of pathologies other than SCI.