Machine learning analysis of cortical activity in visual associative learning tasks with differing stimulus complexity.

Abstract

Associative learning tests are cognitive assessments that evaluate the ability of individuals to learn and remember relationships between pairs of stimuli. The Rutgers Acquired Equivalence Test (RAET) is an associative learning test that utilizes images (cartoon faces and colored fish) as stimuli. RAET exists in various versions that differ in the degree of the complexity of the stimuli used in the given version. It has been observed that differences in stimulus complexity can lead to marked differences in test performance, but the related cortical functional differences remain to be elucidated. In the present study, we introduce a Machine Learning- and Independent Component Analysis-based EEG signal processing pipeline, which can detect such differences. RAET and its reduced stimulus complexity variant, Polygon was administered to 32 healthy volunteers and EEG recordings were made with a 64-channel system. The most remarkable differences between RAET and Polygon were detected in the frontal regions, which can be connected to decision making. On the other hand, the parietal regions showed the lowest number of differences between RAET and Polygon. Some task-related activity in the temporo-occipital region was identified, which shows different dynamics depending on visual stimulus complexity.