Real-Time Decoding of Attention from EEG Using Machine Learning for Adaptive Neurofeedback

This study investigated whether closed-loop EEG neurofeedback could effectively improve sustained attention performance. By training a machine learning classifier to decode in-the-zone and out-of-the-zone attentional states based on reaction time variation, I implemented a real-time feedback system whereby a continuous performance task became easier when participants were decoded to be in-the-zone and harder when they were decoded to be out-of-the-zone. While time in-the-zone based on EEG was not increased, there is evidence that task performance was better in the treatment condition than a sham control feedback condition. Further work is needed to validate this result and identify more definitively the mechanism of this effect.