Abstract
Modulation of sensorimotor rhythms (SMR) was suggested as a control signal for brain-computer interfaces (BCI). Yet, there is a population of users estimated between 10 to 50% not able to achieve reliable control and only about 20% of users achieve high (80–100%) performance. Predicting performance prior to BCI use would facilitate selection of the most feasible system for an individual, thus constitute a practical benefit for the user, and increase our knowledge about the correlates of BCI control. In a recent study, we predicted SMR-BCI performance from psychological variables that were assessed prior to the BCI sessions and BCI control was supported with machine-learning techniques. We described two significant psychological predictors, namely the visuo-motor coordination ability and the ability to concentrate on the task. The purpose of the current study was to replicate these results thereby validating these predictors within a neurofeedback based SMR-BCI that involved no machine learning.Thirty-three healthy BCI novices participated in a calibration session and three further neurofeedback training sessions. Two variables were related with mean SMR-BCI performance: (1) a measure for the accuracy of fine motor skills, i.e., a trade for a person’s visuo-motor control ability; and (2) subject’s “attentional impulsivity”. In a linear regression they accounted for almost 20% in variance of SMR-BCI performance, but predictor (1) failed significance. Nevertheless, on the basis of our prior regression model for sensorimotor control ability we could predict current SMR-BCI performance with an average prediction error of M = 12.07%. In more than 50% of the participants, the prediction error was smaller than 10%. Hence, psychological variables played a moderate role in predicting SMR-BCI performance in a neurofeedback approach that involved no machine learning. Future studies are needed to further consolidate (or reject) the present predictors.
Highlights
Prediction of behavior, performance or decisions of individuals or groups is a popular theme in modern psychology research
The purpose of the current study was to replicate these results thereby validating these predictors within a neurofeedback based sensorimotor rhythms (SMR)-braincomputer interfaces (BCI) that involved no machine learning.Thirty-three healthy BCI novices participated in a calibration session and three further neurofeedback training sessions
In Hammer et al (2012) we argued that the small number of significant psychological predictors was owed to the machine learning approach to BCI control that relies mainly on pattern recognition, and less on human learning
Summary
Prediction of behavior, performance or decisions of individuals or groups is a popular theme in modern psychology research. PREVIOUS STUDY TO INVESTIGATE PREDICTORS In an extensive bi-center study, we investigated whether psychological and physiological parameters would predict SMR-BCI performance based on the Berlin Brain-Computer Interface (BBCI; Blankertz et al, 2010; Hammer et al, 2012), a so called machine learning approach that provides BCI control during the first session after a 30 min calibration period (Blankertz et al, 2007) Since those results serve as a basis for the current study, we present a summery about the methods, results and implications: Eighty healthy participants performed a motor imagery task, first during calibration and subsequently in three feedback sessions, during which they had to operate a one-dimensional (1D) cursor. We explored whether further psychological variables from the test-battery compiled by Hammer et al (2012) would predict SMR-BCI based neurofeedback results
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
