Decoding laser-induced somatosensory information from EEG

Abstract

Recently, our research group has proposed a new way of providing a non-nociceptive tactile sensation with laser [1]. In this study, we aimed to investigate laser-induced somatosensory information represented in cortical activity using the human EEG. The EEG data were acquired using the V-Amp amplifier (Brain Products GmbH, Gilching, Germany) with 16 wet electrodes that were placed on the scalp following the international 10–20 system. Twenty one subjects participated in the study (7 female and mean age of 22.4 years). During the experiment, a mechanical stimulus, a laser stimulus and a heat stimulus were given in a random order to subjects sixty times per stimulus. Subjects described the feeling of laser stimulation as non-painful sensation, painful sensation and no sensation. As described in the previous study, 56.3, 12.3 and 31.4 % of the subjects reported laser stimulation as non-painful, painful and no sensation, respectively [1]. To examine similarity of cortical activity in response to different stimuli, we employed a decoding analysis of the EEG data. In the decoding analysis, we used the linear discriminant analysis (LDA) method to classify the beta (21–28 Hz) event-related desynchronization/synchronization (ERD/S) patterns of EEG into one of the two classes representing every pair of stimuli (a total of six pairs from four stimuli) [2]. Classification error indicated how similar beta ERD/S patterns were between two stimuli: a larger error reflected more difficulty in discriminating patterns and consequently a greater similarity between patterns. The beta ERD/S patterns were estimated using the short time Fourier transform. Baseline correction was implemented using the 0.5 s period before stimulus onset. For each pair of stimuli, one-way ANOVA was used to select four channels that exhibited the most differences in beta ERD/S patterns between classes and classification accuracy was assessed by the leave-four-out cross validation [3] (see Fig. 69 for the classification error between every stimulus pair). The classification results showed that to the beta ERD/S pattern induced by mechanical stimulation, the pattern by non-painful laser stimulation was most similar. Also, the results indicated closeness of cortical activities between non-painful and painful laser stimulations as well as painful laser and thermal stimulations (see Fig. 69). These results suggest that laser might induce similar beta responses whether it evoked painful or non-painful feelings but non-painful laser might share presumably non-nociceptive somatosensory information with mechanical stimulation whereas painful laser shared presumably nociceptive somatosensory information with thermal stimulation. We expect that further information theoretical analyses may reveal more details about somatosensory information encoded in cortical rhythms induced by laser.