Investigating the neural correlates of stroop effect using the multilayer perceptron neural network

Abstract

Aim: The Stroop task, specifically related to semantic conflict processing, is one of the most common cognitive tests examining executive functions. This study aimed to investigate neural correlates of the Stroop interference effect by means of simultaneous electroencephalography (EEG) and functional near infrared spectroscopy (fNIRS) measurements using a machine learning approach.

Methods: A total of forty-five healthy male university students were included in the study. We measured brain activation with EEG/fNIRS systems during the color-word matching Stroop task. Linear and non-linear dynamics of EEG were computed over five frequency sub-bands. fNIRS analysis was conducted with a general linear model. We combined features from both modalities and employed the Multilayer Perceptron (MLP) algorithm to classify incongruent and neutral trials. The Stroop effect in the subregions of the prefrontal cortex was also investigated using statistical analyses.

Results: The results indicated that brain activation due to Stroop interference increased with incongruent stimuli, particularly in the right dorsolateral prefrontal cortex. The Stroop effect was associated with the fractal dimension and power spectral density of EEG. There was a significantly longer reaction time and more task error with incongruent stimuli than neutral trials. MLP classified incongruent and neutral trials with an accuracy rate of 73.3%.

Conclusions: This study is the first to examine the Stroop effect using a multimodal EEG/fNIRS system and machine learning approach. Our study revealed that a hybrid EEG/fNIRS system is an effective neuroimaging tool to study neural correlates of Stroop interference. These findings could be used in future neurological and psychiatric research.