Detection of movement intention from single-trial movement-related cortical potentials using random and non-random paradigms

Abstract

Detection of motor intention with short latency from scalp electroencephalography (EEG) is essential for the development of brain-computer interface (BCI) systems for neuromodulation. This latency determines the temporal association between motor intention and the triggered afferent neurofeedback. In this study, we compared two typical experimental paradigms for the detection of movement intention from EEG. A template-matching algorithm was used to detect movement-related cortical potentials (MRCPs) in eight healthy subjects for two types of cued motor imageries using either a random or non-random cue. For the random cue, the true positive rates of detection of movement intention were 63.5 ± 5.9% (foot movement) and 61 ± 6.5% (right hand movement). Detection occurred 102.8 ± 119.3 ms and 112.2 ± 104 ms prior to onset of execution cue. On the other hand, foot and hand movement intentions were detected significantly earlier (p < 0.05) (198 ± 147.3 ms and 206 ± 134.2 ms prior to onset, respectively) and with a greater true positive rate (p < 0.05) (75.3 ± 5.5% and 70.2 ± 6.1%) when non-random cues were used. These results suggest that a non-random cue paradigm is preferable to a typical random cue in BCI systems designed for neuromodulation. However, the important consideration of variable practice afforded by the random cue, which is known to facilitate learning, requires further study.