EEG-GAN: Generative adversarial networks for electroencephalograhic (EEG) brain signals

By Kay Gregor Hartmann, Robin Tibor Schirrmeister, Tonio Ball. 2018

286 citations recorded by the source corpus.

Generative adversarial networks (GANs) are recently highly successful in generative applications involving images and start being applied to time series data. Here we describe EEG-GAN as a framework to generate electroencephalographic (EEG) brain signals. We introduce a modification to the improved training of Wasserstein GANs to stabilize training and investigate a range of architectural choices critical for time series generation (most notably up- and down-sampling). For evaluation we consider and compare different metrics such as Inception score, Frechet inception distance and sliced Wasserstein distance, together showing that our EEG-GAN framework generated naturalistic EEG examples. It thus opens up a range of new generative application scenarios in the neuroscientific and neurological context, such as data augmentation in brain-computer interfacing tasks, EEG super-sampling, or restoration of corrupted data segments. The possibility to generate signals of a certain class and/or with specific properties may also open a new avenue for research into the underlying structure of brain signals.

Subfields: Cognitive Neuroscience

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

1. Chin-Teng Lin
2. Peng Xu
3. Alireza Jolfaei
4. Xiaotong Gu
5. Zehong Cao
6. Tzyy-Ping Jung
7. Dongrui Wu
8. Dacheng Tao

Citations in the corpus

• Generative Adversarial Networks (2021, neutral): “By providing a thorough overview of GANs, we hope to make this complex and challenging topic more accessible to researchers and practitioners in the field, and inspire new innovations and applications.”
• Data Augmentation for Deep-Learning-Based Electroencephalography. (2020, methodology): “The class conditioned implementation allowed the generator and discriminator to avoid mode collapsing, which is responsible for trapping the data generated from the GAN in some specific modes [65].”
• Data Augmentation techniques in time series domain: a survey and taxonomy (2022, background): “The model G generates samples trying to replicate, without copying, the distribution, while the model D discriminates the real samples from the fake samples.”
• Deep learning-based electroencephalography analysis: a systematic review (2019, methodology): “More recently, studies such as [75, 183] achieved promising results without the need to extract features.”
• Synthesis of Realistic ECG using Generative Adversarial Networks (2019, methodology): “GANs were also used to generate single-channel electroencephalogram (EEG) data for motor movement in both the left and right hand [25].”
• Generative Adversarial Networks: A Survey and Taxonomy (2019, background): “GANs have been applied to some other fields, (e.g., time series generation [20], [21], [103] and natural language processing [15], [104]–[106]) with some success.”
• Generative Adversarial Networks in Time Series: A Systematic Literature Review (2022, unknown): “Medical/Physiological Generation LSTM-LSTM, [31], [77], [2], [44], [45], [97] LSTM-CNN, [11] [21] BiLSTM-CNN, [111] BiGridLSTM-CNN, [49] CNN-CNN, [47], [33] AE-CNN, [81] FCNN [104] EEG, ECG, EHRs, PPG, EMG, Speech, NAF, MNIST, Synthetic sets TSTR, MMD, Reconstruction error, DTW, PCC, IS, FID, ED, SW”
• A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications (2020, background): “Additionally, GANs have made achievements in sequential data tasks, such as those involving natural language, music, speech, voice [268], [269], and time series data [270], [271], [272], [273].”
• Quick and Easy Time Series Generation with Established Image-based GANs (2019, unknown): “elated Work Few studies have used GANs to produce time series data as they have mainly been developed for the generation of images. Some recent results showed promise in synthesising time series data [2,6]. Hartmann et al. [6] used channel FCC4h recorded from a 128-electrode electroencephalograph (E”
• EEG-Based Brain-Computer Interfaces (BCIs): A Survey of Recent Studies on Signal Sensing Technologies and Computational Intelligence Approaches and Their Applications (2020, background): “ples to mimic real-world data via generative models so that the training data sample number can be increased [122].”