Visual Search Diffculty Modulates the Functional Relevance and Lateralization of Cortical Regions

The brain prioritizes the allocation of its limited cognitive resources to a small number of regions that are functionally relevant to the ongoing task, thereby facilitating efficient performance. Visual search tasks are no exception, as they likewise involve a functionally driven and competitive allocation of cognitive resources. Using task-based functional magnetic resonance imaging (ts-fMRI), this study investigated the functional relevance of cortical regions during goal-directed visual search to gain deeper insight into the execution of this top-down process. Herein, we propose an analytical framework that integrates a Multi-Scale Short-Term Attention Network (MSTANet) with a “mask-one-ROI-off” loop perturbation evaluation to investigate the potential of deep learning–based multivariate pattern analysis (MVPA). Based on the Automatic Anatomical Labeling (AAL) atlas, MSTANet achieved an average pattern recognition accuracy of up to 0.839 in classifying ts-fMRI time courses across two levels of visual search difficulty and a baseline condition. The “mask-one-ROI-off” loop perturbation evaluation quantified the informational value of individual regions of interest (ROIs) for pattern recognition, identifying cortical regions such as the supramarginal gyrus (SMG) and superior occipital gyrus (SOG) as highly functionally relevant to visual search. Furthermore, we examined the functional lateralization of these task-relevant ROIs across attention and visual networks, as well as their modulation under varying levels of endogenous attentional demand.