Classifying attentional dynamics from EEG signals ::feature based perceptual attentional control

The general goal of this project is to understand if something as generic as paying attention to visual stimuli, whether intended or driven by distractor objects, can be classified well. People’s ability to behave effectively in everyday situations is critically dependent on “selective attention”, which is the ability to promote the processing of objects that match our current behavioral goals and suppress those objects that do not match those goals. The last decades have provided significant advances in terms of brain and cognitive mechanisms orchestrating selective attention as well as their role in enhancing perception and supporting the learning of new information. However, this knowledge of both processes of intended and distracted attention has yet to be discriminated. Goal: If distraction and intention can be classified by only neural input, particularly in real-world, multisensory environments, this could bring forth knowledge of the underlying differences in mechanisms additive to the traditional neural analysis methods. For this reason, it is particularly useful to model and understand cognition processes through statistical modelling, such as those in Machine Learning (ML) applications on attentional control data. Methodology: In particular, employingML techniques to discriminate between visual selective attention to distractor objects vs. intended object, in an unbiased and automatic manner, without relying on subjective evaluation. Results: In this endeavor, a linear classifier was trained to successfully classify attention to distractor object, intended object, or neither, with the best accuracy score of 0.65 (chance accuracy score = 0.33). Additionally, the selection of features from N2PC regions resulted in the best accuracy score of 0.65 while decreasing feature size to 10.9 perfect of total features (14/128 electrodes). However, the non-N2PC region features suggested that attention is a process that uses whole-brain activity