.png){kind=icon}
.png){kind=icon}
Return
Cortical alpha changes during visuospatial attention: a deep learning-enriched EEG analysis
.png){kind=icon}
OA
AI
DOI：10.1093/cercor/bhag026.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Covert visuospatial attention in anticipation of a stimulus is known to topographically modulate alpha-band (8 to 14 Hz) brain activity. However, the specific cortical regions involved remain unclear. Here, we conducted a whole-cortex analysis of alpha-band changes by examining source-level electroencephalographic (EEG) signals during a cued visuospatial attention task via a novel approach integrating conventional alpha power analysis with a deep learning technique based on an interpretable convolutional neural network (CNN). Conventional metrics contrasting alpha power between leftward and rightward attention suggested a robust, selective involvement of the left parietal lobe (superior and inferior parietal cortices) and a broader involvement of right hemisphere regions with less parietal contribution. The CNN-based approach, which discriminated the attention direction from the source-level EEG signals and identified the most discriminative regions in alpha band, refined these findings, corroborating the dominant role of the left parietal lobe and limited involvement of the right parietal lobe restricted to the supramarginal gyrus. The obtained findings are interpreted in terms of a more tonic engagement (disinhibition) of right parietal lobe, leaving less dynamic range for condition-dependent alpha modulation. This study not only improves the characterization of alpha-band attention-related changes but also presents a novel combined approach to investigate brain oscillations.
Keywords:
anticipatory visuospatial attention
alpha power
EEG source reconstruction
interpretable convolutional neural network
parietal cortex