Detection of movement intention from movement-related cortical potentials with different paradigms

Abstract

In this study, we compared the effects of two imagery paradigms typically used within the field of brain computer interfaces on the detection of movement intention from scalp electroencephalography (EEG). This issue is important in the rehabilitation area because of its direct relation with appropriately timed neurofeedback. Subjects were asked to imagine hand or foot movements using either a random or a non-random cue. Templates were constructed individually for each subject. Movement intent was detected according to the correlation between the movement related cortical potentials (MRCP) of single trials with the initial part of the template. The large Laplacian filter was used to increase the signal to noise ratio (SNR). For the random cue, the true positive rate (TPR) of detection of movement intention was 63.5±5.9% for foot movement and the corresponding detection latency was 202.8±129.5 ms before movement onset. For the non-random cue, foot movement intention was detected with TPR of 75.3±5.5% and latency of 291±169.3 ms. These results demonstrate that cue type, random or non-random, has a significant effect on the performance of MRCPbased movement intention detection algorithms.