Generative adversarial networks in EEG analysis: an overview

Abstract

Electroencephalogram (EEG) signals have been utilized in a variety of medical as well as engineering applications. However, one of the challenges associated with recording EEG data is the difficulty of recording large amounts of data. Consequently, data augmentation is a potential solution to overcome this challenge in which the objective is to increase the amount of data. Inspired by the success of Generative Adversarial Networks (GANs) in image processing applications, generating artificial EEG data from the limited recorded data using GANs has seen recent success. This article provides an overview of various techniques and approaches of GANs for augmenting EEG signals. We focus on the utility of GANs in different applications including Brain-Computer Interface (BCI) paradigms such as motor imagery and P300-based systems, in addition to emotion recognition, epileptic seizures detection and prediction, and various other applications. We address in this article how GANs have been used in each study, the impact of using GANs on the model performance, the limitations of each algorithm, and future possibilities for developing new algorithms. We emphasize the utility of GANs in augmenting the limited EEG data typically available in the studied applications.

Improving cross-subject classification performance of motor imagery signals: a data augmentation-focused deep learning framework

“…Based on this notion, other types of well-known AEs, such as stacked AE [30] and variational AE (VAE) [31] have been derived. In recent years, there has been a growing interest in utilizing GANs for EEG data augmentation [32], inspired by their well-established success in image and sound data synthesis. However, generating synthetic EEG signal sequences using GANs necessitates the development of specialized techniques and raises certain concerns that need to be addressed.…”

Applications for Deep Learning in Epilepsy Genetic Research

“…Data augmentation has been successfully applied to epileptic seizure prediction from EEG data [21], and to delineate the seizure-onset zone in individuals with focal epilepsy [86]. See Habashi, Azab, et al [87] for a full review. In genomics, Wei, Li, et al [84] used data augmentation of genetic cancer data to demonstrate how this strategy improves cancer classification.…”

A magnetoencephalography dataset during three-dimensional reaching movements for brain-computer interfaces

“…Although our previous studies utilizing the shared data revealed the neural mechanisms 1 , 5 , 6 and successfully trained the deep learning algorithms 6 , 22 , the dataset may not be sufficient for training super-giant AI models like the generative pre-trained transformer (GPT). The issue of data scarcity can be partially overcome by data augmentation 33 , 34 , generation 35 , and transfer learning 36 techniques. Data augmentation involves slight modifications to the existing data to increase its quantity, achieved through methods such as noise addition, sliding windows, and recombination of segmentation.…”