Generative Representation of Aggregate Brain Activity: A Deep Autoencoder Approach for EEG Topoplot Summarization

This research aims to assess a Deep Convolutional Autoencoder (CAE) framework for representative EEG topoplot summarization using latent space aggregation. In order to produce representative EEG topoplot summaries while maintaining important spatial features, we suggest a Deep Convolutional Autoencoder (CAE) with latent space aggregation. Prior to group-level aggregation and image reconstruction, EEG topoplots are simplified into latent representations that resemble baseline artifacts. An adolescent EEG dataset obtained during a Go/No-Go Association Task involving addiction stimuli was used to test our methodology. The frontal-temporal predominance of normal respondents and the prominent temporal-occipital activation of at-risk respondents, primarily in those with slower responses, are caused by distinct activation patterns that are associatively aroused by attentional and memory bias. These results support the use of secure EEG topoplot summarization in addiction research using CAE-based latent space aggregation.