Exploring Deep Learning Models for EEG Neural Decoding

2025-03-20

AI Summary

Decoding high-level semantic features—specifically, biological versus non-biological object distinctions—from electroencephalography (EEG) signals remains challenging due to the low signal-to-noise ratio and complex spatiotemporal dynamics of neural responses. Method: Leveraging the large-scale THINGS EEG dataset (46 participants), we systematically benchmarked 15 deep learning architectures—including CNNs, RNNs, Transformers, MLPs, and hybrid models—for temporal neural decoding, comparing them against state-of-the-art linear decoders. Contribution/Results: Linear models consistently failed, whereas all deep nonlinear models significantly outperformed the baseline (mean accuracy improvement of 12.7%), providing the first empirical evidence for both the efficacy and necessity of deep learning in modeling high-order cognitive neural representations. Notably, several models exhibited interpretable, fine-grained category-selective neural response patterns aligned with semantic object distinctions. This work establishes a new paradigm for computational modeling of brain mechanisms underlying semantic perception.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Neural decoding is an important method in cognitive neuroscience that aims to decode brain representations from recorded neural activity using a multivariate machine learning model. The THINGS initiative provides a large EEG dataset of 46 subjects watching rapidly shown images. Here, we test the feasibility of using this method for decoding high-level object features using recent deep learning models. We create a derivative dataset from this of living vs non-living entities test 15 different deep learning models with 5 different architectures and compare to a SOTA linear model. We show that the linear model is not able to solve the decoding task, while almost all the deep learning models are successful, suggesting that in some cases non-linear models are needed to decode neural representations. We also run a comparative study of the models' performance on individual object categories, and suggest how artificial neural networks can be used to study brain activity.

Problem

• Decoding high-level object features from EEG data using deep learning models
• Comparing performance of 15 deep learning models versus a linear model
• Investigating artificial neural networks for studying brain activity patterns

Innovation

• Deep learning models decode EEG neural activity
• Non-linear models outperform linear decoding
• Comparative study of 15 deep architectures