Learning Data Augmentation for Brain Tumor Segmentation with Coarse-to-Fine Generative Adversarial Networks

The high human labor demand involved in collecting paired medical imaging data severely impedes the application of deep learning methods to medical image processing tasks such as tumor segmentation. The situation is further worsened when collecting multi-modal image pairs. However, this issue can be resolved through the help of generative adversarial networks, which can be used to generate realistic images. In this work, we propose a novel framework, named TumorGAN, to generate image segmentation pairs based on unpaired adversarial training. To improve the quality of the generated images, we introduce a regional perceptual loss to enhance the performance of the discriminator. We also develop a regional L1 loss to constrain the color of the imaged brain tissue. Finally, we verify the performance of TumorGAN on a public brain tumor data set, BraTS 2017. The experimental results demonstrate that the synthetic data pairs generated by our proposed method can practically improve tumor segmentation performance when applied to segmentation network training.

Download Full-text

Brain Tumor Segmentation Using 3D Generative Adversarial Networks

International Journal of Pattern Recognition and Artificial Intelligence ◽

10.1142/s0218001421570020 ◽

2020 ◽

pp. 2157002

Author(s):

Yitong Li ◽

Yue Chen ◽

Y. Shi

Keyword(s):

Brain Tumor ◽

Treatment Options ◽

Medical Image Analysis ◽

Three Dimensional ◽

Subsequent Treatment ◽

Generative Adversarial Networks ◽

Tumor Segmentation ◽

High Morbidity ◽

Brain Tumor Segmentation ◽

Adversarial Networks

Brain tumors have high morbidity and may lead to highly lethal cancer. In clinics, accurate segmentation of tumors is the means for diagnosis and determination of subsequent treatment options. Due to the irregularity and blurring of tumor boundaries, accurately segmenting the tumor lesions has received extensive attention in medical image analysis. In view of this situation, this paper proposed a brain tumor segmentation method based on generative adversarial networks (GANs). The GAN architecture consists of a densely connected three-dimensional (3D) U-Net used for segmentation and a classification network for discrimination, both of which use 3D convolutions to fuse multi-dimensional context information. The densely connected 3D U-Net model introduces a dense connection to accelerate network convergence, extracting more detailed information. The adversarial training makes the distribution of segmentation results closer to that of labeled data, which enables the network to segment some unexpected small tumor subregions. Alternately, train two networks and finally achieve a highly accurate classification of each voxel. The experiments conducted on BraTS2017 brain tumor MRI dataset show that the proposed method has higher accuracy in brain tumor segmentation.

Download Full-text

Data Augmentation for Brain-Tumor Segmentation: A Review

Frontiers in Computational Neuroscience ◽

10.3389/fncom.2019.00083 ◽

2019 ◽

Vol 13 ◽

Cited By ~ 8

Author(s):

Jakub Nalepa ◽

Michal Marcinkiewicz ◽

Michal Kawulok

Keyword(s):

Brain Tumor ◽

Data Augmentation ◽

Tumor Segmentation ◽

Brain Tumor Segmentation

Download Full-text

Improving Brain Tumor Segmentation with Data Augmentation Strategies

10.21467/proceedings.114.2 ◽

2021 ◽

Author(s):

Radhika Malhotra ◽

Jasleen Saini ◽

Barjinder Singh Saini ◽

Savita Gupta

Keyword(s):

Brain Tumor ◽

Data Augmentation ◽

Large Data ◽

Training Data ◽

Tumor Segmentation ◽

Biomedical Image Processing ◽

Brain Tumor Segmentation ◽

Biomedical Image ◽

Augmentation Strategies ◽

Unseen Data

In the past decade, there has been a remarkable evolution of convolutional neural networks (CNN) for biomedical image processing. These improvements are inculcated in the basic deep learning-based models for computer-aided detection and prognosis of various ailments. But implementation of these CNN based networks is highly dependent on large data in case of supervised learning processes. This is needed to tackle overfitting issues which is a major concern in supervised techniques. Overfitting refers to the phenomenon when a network starts learning specific patterns of the input such that it fits well on the training data but leads to poor generalization abilities on unseen data. The accessibility of enormous quantity of data limits the field of medical domain research. This paper focuses on utility of data augmentation (DA) techniques, which is a well-recognized solution to the problem of limited data. The experiments were performed on the Brain Tumor Segmentation (BraTS) dataset which is available online. The results signify that different DA approaches have upgraded the accuracies for segmenting brain tumor boundaries using CNN based model.

Download Full-text