GAN Augmentation: Augmenting Training Data using Generative Adversarial Networks

By Christopher Bowles, Liang Chen, Ricardo Guerrero, Paul Bentley, Roger N. Gunn, Alexander Hammers, D. A. Dickie, M. Hernández, Joanna M. Wardlaw, Daniel Rueckert. 2018

497 citations recorded by the source corpus.

One of the biggest issues facing the use of machine learning in medical imaging is the lack of availability of large, labelled datasets. The annotation of medical images is not only expensive and time consuming but also highly dependent on the availability of expert observers. The limited amount of training data can inhibit the performance of supervised machine learning algorithms which often need very large quantities of data on which to train to avoid overfitting. So far, much effort has been directed at extracting as much information as possible from what data is available. Generative Adversarial Networks (GANs) offer a novel way to unlock additional information from a dataset by generating synthetic samples with the appearance of real images. This paper demonstrates the feasibility of introducing GAN derived synthetic data to the training datasets in two brain segmentation tasks, leading to improvements in Dice Similarity Coefficient (DSC) of between 1 and 5 percentage points under different conditions, with the strongest effects seen fewer than ten training image stacks are available.

Subfields: Computer Vision and Pattern Recognition, Generative Models

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Generative Models

Citations in the corpus

Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules (2019, background): “Augmentation has been shown to have a large impact on image modalities where data is scare or expensive to generate, like medical imaging (Bowles et al., 2018; Frid-Adar et al., 2018) or non-supervised learning approaches (Mundhenk et al., 2017).”
GAN Memory with No Forgetting (2020, background): “GAN [25] setup as the key component of that generative memory, which we term GAN memory, because (i) GANs have shown remarkable power in synthesizing realistic high-dimensional samples [9, 51, 33, 34]; (ii) by modeling the generative process of training data, GANs summarize the data statistical info”
Generative Adversarial Network in Medical Imaging: A Review (2018, background): “This is achieved by isolating the latent encoding of AD and performing arithmetic operation in the latent space (Bowles et al., 2018b).”
Global and Local Contrastive Self-Supervised Learning for Semantic Segmentation of HR Remote Sensing Images (2021, background): “To solve the problem of an insufficient number of labeled samples, one strategy is to generate more samples through data augmentation [27], generative adversarial networks (GANs) [28], etc.”
On Leveraging Pretrained GANs for Limited-Data Generation (2020, methodology): “For the problem of limited-data generation, to naively train a GAN on the limited data is prone to overfitting for powerful GAN models with a lot of parameters, which are essential for realistic generation (Bermudez et al., 2018; Bowles et al., 2018; Frid-Adar et al., 2018; Finlayson et al., 2018).”
Imbalance Problems in Object Detection: A Review (2019, methodology): “Many studies [156], [157], [158] have successfuly used GANs to generate examples for under-represented classes for various image classification problems.”
Artificial intelligence and machine learning for medical imaging: A technology review. (2021, unknown): “…we believe that GAN-based models have the potential to better sample the whole data distribution and generate more realistic images than traditional approaches (e.g. rotation, flipping, etc), which may contribute to a higher generalizability of the model [188] and more efficient training [195].”
A Review of Deep Learning Models for Time Series Prediction (2019, background): “Considering the capability of generating missing data points [192], the prediction for incomplete time series is a possible future topic deserving further study.”
Data Augmentation with Manifold Exploring Geometric Transformations for Increased Performance and Robustness (2019, methodology): “Attempts have been made to solve the GAN labelling problem in the medical context [8], by generating Brain CT scans along with a paired segmentation label map.”
Vision Transformers for Remote Sensing Image Classification (2021, background): “One interesting approach for data augmentation is the one based on generative models, in which models such as GANs [50] learn the distribution of the data to create synthetic samples that are as similar as possible to the images drawn from the original dataset.”
Controlling generative models with continuous factors of variations (2020, background): “More control could be used to improve existing approaches which aim at generating new training examples (Bowles et al., 2018) by allowing the user to choose more speciﬁc properties of the generated images.”
On Leveraging Pretrained GANs for Generation with Limited Data (2020, methodology): “…zooming, rotation, or flipping, GANs trained on limited data can be leveraged for synthetic augmentations, like synthesized observations with transformed styles (Wang & Perez, 2017) or fake observation-label/segmentation pairs (Bowles et al., 2018; Frid-Adar et al., 2018; Han et al., 2019; 2020).”
Generalizing Deep Learning for Medical Image Segmentation to Unseen Domains via Deep Stacked Transformation (2020, background): “In speciﬁc applications of medical image segmentation, image rotation and GAN-based augmentations have been shown to improve the performance in 2D data for both CT and MRI [27], as they can extrapolate and interpolate the manifold of data, respectively.”
Contrastive learning of global and local features for medical image segmentation with limited annotations (2020, methodology): “Augmentation through random afﬁne [13], elastic [49] and contrast transformations [29, 47], Mixup [61, 19] and synthetic data generation via GANs [22, 15, 52, 8] have been explored.”