A new approach for single-trial detection of laser-evoked potentials and its application to pain prediction

Abstract

Single-trial detection of evoked brain potentials is essential for many research topics in neural engineering and neuroscience. In present study, a novel approach, which combines common spatial pattern (CSP) and multiple linear regression (MLR), is proposed to for single-trial detection of pain-related laser-evoked potentials (LEPs). The CSP method is effective in separating laser-evoked EEG response from ongoing EEG activity, while MLR makes an automatic and reliable estimation of the amplitudes and latencies of N2 and P2 from single-trial LEP waveforms. The MLR coefficients are further used for the prediction of pain perception, which is of great importance for both basic and clinical applications. The prediction is performed with both binary (classification of low pain and high pain) and continuous (regression on a continuous scale from 0 to 10) outcomes. The results show that the proposed methods could provide reliable performance at both with- and cross-individual levels.