Sustained Attention States Recognition with EEG and Eye-Tracking in the GradCPT

The objective of the current study is to explore the feasibility of online recognition of human sustained attention states using electroencephalography (EEG) and eye-tracking technology in the gradual onset continuous performance task (gradCPT). Sixteen volunteer participants each completed a 2-min practice session and three 8-min experimental sessions of gradCPT. EEG and eye-tracking data were collected during the experimental sessions. Six machine learning algorithms, including logistic regression (LR), linear discriminant analysis (LDA), support vector machines (SVM), random forest (RF), k-nearest neighbors (kNN), and artificial neural networks (ANN), were tested in their performance in recognizing in-the-zone and out-of-the-zone periods. On the behavioral level, the results were consistent with the previous gradCPT studies. Among the machine learning algorithms, SVM and LR yielded above-average performance, with a classification accuracy of 0.62; SVM was the best performer considering balanced sensitivity and specificity. This study demonstrated that it is feasible to detect human sustained station states using frontal-channels EEG and eye-tracking features with above-chance accuracy.