Predictive models of the EEG activity based on eye movements during reading

Objective. This study aimed to explore the interrelationship between eye movements and EEG activity and determine which measure has better quality to predict reading efficiency. Methods. The study involved measuring eye movements and EEG from a limited number of leads while 40 healthy adults were reading a novel fragment. Machine learning models (CatBoost, KNN, linear regression) were then applied to predict EEG from eye movements, eye movements from EEG, and reading efficiency from both biological sources. Results. Throughout the study, CatBoost showed the best predictive quality. Of the EEG activity, the alpha-activity was the most accurately predicted from eye movements. Theta- and alpha-power were best predicted in posterior leads, while beta-power and engagement indices were best predicted in frontal leads. Then we predicted eye movements based on the EEG activity, the fixation number being the best-predicted feature. Finally, the EEG activity predicted reading efficiency much better than the eye movements data. Discussion. We describe our data in light of previous EEG experiments highlighting the role of posterior alpha and frontal beta bands in attention and reading processes. Conclusions. EEG activity and eye movements have been revealed to be interrelated, that is future studies can reduce the number of measurements in field settings, where limited technical equipment is available.