Abstract:Most models in cognitive and computational neuroscience trained on one subject do not generalize to other subjects due to individual differences. An ideal individual-to-individual neural converter is expected to generate real neural signals of one subject from those of another one, which can overcome the problem of individual differences for cognitive and computational models. In this study, we propose a novel individual-to-individual EEG converter, called EEG2EEG, inspired by generative models in computer vision. We applied THINGS EEG2 dataset to train and test 72 independent EEG2EEG models corresponding to 72 pairs across 9 subjects. Our results demonstrate that EEG2EEG is able to effectively learn the mapping of neural representations in EEG signals from one subject to another and achieve high conversion performance. Additionally, the generated EEG signals contain clearer representations of visual information than that can be obtained from real data. This method establishes a novel and state-of-the-art framework for neural conversion of EEG signals, which can realize a flexible and high-performance mapping from individual to individual and provide insight for both neural engineering and cognitive neuroscience.

What problem does this paper attempt to address?

The paper aims to address the issue of model generalization caused by individual differences in the fields of cognitive neuroscience and computational neuroscience. Specifically, due to significant differences in neural activity between different individuals, models trained on a single subject are difficult to generalize to other subjects. To solve this problem, the study proposes a new inter-individual EEG converter—EEG2EEG. ### Main Contributions: 1. **Proposed EEG2EEG Model**: Based on the U-Net structure in Generative Adversarial Networks (GAN), this model can convert the EEG signals of one individual into the EEG signals of another individual, thereby overcoming the problem of individual differences. 2. **High Conversion Performance**: Experimental results show that EEG2EEG can effectively learn the mapping of EEG signals from the source individual to the target individual, and the generated EEG signals are highly similar to the real EEG signals. 3. **Information Representation Capability**: Through Multivariate Pattern Analysis (MVPA), it was found that the generated EEG signals not only contain visual information comparable to real data but also, in some cases, even outperform real data, demonstrating a certain denoising capability. 4. **Flexibility and Application Prospects**: The EEG2EEG model can not only reduce the time and cost of ERP research but also potentially be applied in brain-to-brain communication and the study of individual difference mechanisms. ### Methods and Experiments: - **Dataset**: The THINGS EEG2 dataset was used for training and testing. - **Model Architecture**: An encoder-decoder structure was adopted, including convolutional layers, skip connections, and a cosine distance loss function. - **Evaluation Metrics**: The similarity between the generated data and the real data was evaluated using metrics such as Spearman correlation coefficient, pattern generation accuracy, and silhouette correlation. - **Comparative Experiments**: Compared with other models (linear model, model without skip connections, model without cosine loss function), the superiority of the EEG2EEG model was demonstrated. In summary, this study proposes a novel and efficient method for inter-individual EEG conversion, providing new perspectives and technical means for future neuroengineering and cognitive neuroscience research.

Generate your neural signals from mine: individual-to-individual EEG converters

Neurocognition-inspired Design with Machine Learning

GET: A Generative EEG Transformer for Continuous Context-Based Neural Signals

Neural decoding from stereotactic EEG: accounting for electrode variability across subjects

Toward reliable signals decoding for electroencephalogram: A benchmark study to EEGNeX

A GAN model encoded by CapsEEGNet for visual EEG encoding and image reproduction

Inter-individual and inter-site neural code conversion without shared stimuli

Generalized neural decoders for transfer learning across participants and recording modalities

Enhancing EEG Signal Generation through a Hybrid Approach Integrating Reinforcement Learning and Diffusion Models

EEG decoding with conditional identification information

Realizing the Application of EEG Modeling in BCI Classification: Based on a Conditional GAN Converter

A Robust 3D-Convolutional Neural Network-Based Electroencephalogram Decoding Model for the Intra-Individual Difference

EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces

Neurodm: Decoding and Visualizing Human Brain Activity with Eeg-Guided Diffusion Model

A shared neural encoding model for the prediction of subject-specific fMRI response

A Novel End-to-End Framework to Image Cortical Networks from EEG

Personalized visual encoding model construction with small data

Generative Adversarial Network (GAN) for Simulating Electroencephalography

Subject-Agnostic Transformer-Based Neural Speech Decoding from Surface and Depth Electrode Signals

EEG2IMAGE: Image Reconstruction from EEG Brain Signals

NT-ViT: Neural Transcoding Vision Transformers for EEG-to-fMRI Synthesis