A electroencephalography (EEG) based braincomputer interface (BCI) provides a communication channel to operate the external environment by decoding brain patterns. Movement-Related Cortical Potentials (MRCPs) are one particular type of EEG pattern during movement of peripheral limbs. The performance of motor imagery (MI) is related to pattern of MRCPs by planning simulation. In resent decade, MI-based BCI have shown potential as its performance significantly improved. In this study, the feasibility of selected-based method was proposed by compared with conventional method. The detection accuracy overall performances were 72.42± 3.12\%. When a top 97.8% trial is selected, overall performance improved approximately 3.15% compared to baseline. When MI were analyzed in non-selected trials, C3 and C4 channel showed no different aspects in left and right class respectively. The brain state was changed after the cue appeared, and these power of delta band appeared in all subjects. The performance of classification was improved by rejecting trials with no difference between the state before and after cue.