Effects of Long-Term Meditation Practices on Sensorimotor Rhythm Based BCI Learning

Abstract

Sensorimotor rhythm (SMR) based brain-computer interfaces (BCIs) provide an alternative pathway for users to perform motor control using motor imagery (MI). Despite the non-invasiveness, ease of use and low cost, this kind of BCI has limitation due to long training times and BCI inefficiency— where a subpopulation cannot generate decodable EEG signals to perform the control task. Meditation is a mental training method to improve mindfulness and awareness, and is reported to have a positive effect on one’s mental state. Here we investigate the behavioral and electrophysiological differences between experienced meditators and meditation naïve subjects in 1-dimensional and 2-dimensional cursor control tasks. We found that within subjects who have room for improvement, meditators outperformed control subjects in both tasks, and there were fewer BCI insufficient subjects in the meditator group. Finally, we also explored the neurophysiological difference between the two groups, and showed that meditators had higher SMR predictor and were better able to generate decodable EEG signals to achieve SMR BCI control.

Immediate effects of short-term meditation on sensorimotor rhythm-based brain–computer interface performance

“…Data from these two runs were not used for further analysis. Given that prior work shows BCI learning can occur even after 20-30 runs (Jiang et al, 2021b), discarding the very first run will have a limited effect on studying the BCI learning in this work, and we can avoid potentially inaccurate BCI data when the subjects are just starting to learn the BCI task. Thus, only 6 runs of BCI data were saved from session 1 part 1.…”

“…The cursor speed was determined by the normalized AR amplitude difference and the cursor position was updated every 40 ms. For horizontal motion (LR tasks), the control signal was produced by taking the difference in AR amplitude between the two electrodes (C4 -C3), and for vertical motion (UD tasks), the signal was produced by taking the sum of the AR amplitudes of the two electrodes (C4 + C3). Each trial had three possible outcomes: a "hit" when the cursor successfully hit the target within 6 s and changed colors; a "miss" when the cursor hit the invisible target on the opposite side from where the target appeared; and an "invalid" when the subject was unable to hit the target with the cursor within 6 s. BCI performance, or MI accuracy, was quantified using percent valid correct (PVC) (Doud et al, 2011;Cassady et al, 2014;Meng et al, 2016;Edelman et al, 2019;Jiang et al, 2021b), which is the proportion of the number of "hits" within the total number of "hits" and "misses. "…”

“…In this particular study, the TMS was used in lieu of the Freiburg Mindfulness Inventory (FMI) and the Mindfulness Attention Awareness Scale (MAAS), which were used in a previous study (Jiang et al, 2021b). Unlike the TMS, the FMI and the MAAS ask the respondent to look back at the past few days to answer the questions (Brown and Ryan, 2003;Walach et al, 2006), thus rendering these surveys inappropriate for this study.…”

“…Another challenge with SMR-based BCI is that it requires a relatively long training time (He et al, 2020). Unlike exogenous BCIs that do not require a lot of training, such as those that use P300 (Eidel and Kübler, 2020) or steady-state visually-evoked potentials (SSVEP) (Chen et al, 2015), SMRbased BCI, which operates using users' volitional CNS signals as input, exhibits a learning curve and requires several sessions of training before users can achieve good performance (Jiang et al, 2021b;Stieger et al, 2021b).…”

Continuous sensorimotor rhythm based brain computer interface learning in a large population

“…One encouraging trend in BCI is to use artificial neural networks to decode brain states 13 , 35 – 37 . Critically, progress in creating robust and generalizable BCI decoding systems is currently hindered by the limited data available to train these decoding models.…”