Predicting Wrist Movement Trajectory from Ipsilesional ECoG in Chronic Stroke Patients

Abstract

Recently, there have been several approaches to utilize a Brain-Computer Interface (BCI) for chronic stroke patients. The prediction of movement trajectory based on recorded brain activity could thereby help to improve BCI-guided stroke rehabilitation or could be used for control of an assistive device, like an orthosis or a robotic arm. One problem in predicting movement trajectory in stroke patients are compensatory movements, which make it difficult to link specific brain activity to movement intention. In this paper we compare different methods for trajectory prediction and show how Canonical Correlation Analysis (CCA) can be used to predict movement trajectories. Based on the results, we argue that the resulting trajectory prediction is closer to the actual movement intention. We further show how the transformation matrices obtained by CCA can be interpreted and discuss how this interpretation might be useful to get information regarding compensatory movements in stroke and the underlying patterns of brain activity.