A Bi-Attention Adversarial Network for Prostate Cancer Segmentation

Abstract

Prostate cancer is one of the most prevalent cancers among men. Early detection of this cancer could effectively increase the survival rate of the patient. In this paper, we propose a Bi-attention adversarial network for the prostate cancer segmentation automatically. The proposed architecture consists of the generator network and discriminator network. The generator network aims to generate the predicted mask of the input image, while the discriminator network aims to further improve the generator performance with adversarial learning by discriminating the generator predicted mask and the true label mask. For better improving the segmentation performance, we combine two attention mechanisms with the generator network to learn more global and local features. Extensive experiments on the T2-weighted (T2W) images have demonstrated our model could achieve state-of-the-art segmentation performance compared with other methods.

Accurate Prostate Cancer Detection and Segmentation Using Non-Local Mask R-CNN With Histopathological Ground Truth

“…Kohl S et al applied an adversarial neural network for IL segmentation, resulting in a DSC of 0.41 ± 0.28; however, MRI-based delineations were used (26). Zhang G et al proposed a bi-attention adversarial network and achieved a DSC of 0.859; however, lesion patches were cropped out before the segmentation (27). In addition, studies on 3D PCa segmentation DL models are limited, the small number of training patients, small volume size of lesions and the low axial resolution poses challenges to the training of a good 3D network.…”

An Improved 2D U-Net Model Integrated Squeeze-and-Excitation Layer for Prostate Cancer Segmentation

“…ere are many research studies [5,6,[10][11][12] took the deep learning method the same with as to achieve prostate cancer segmentation on MRI because it comes to more remarkable performance in the field compared to the traditional method. e idea of making an optimisation based on U-net has attracted much attention in recent years; many related research studies have made good results.…”

“…e application of the SE layer took much inspiration from the channel attention utilized in a biattention adversarial network designed by Zhang et al [12], which proves to have a positive effect on improving model performance.…”

A Multi-Channel and Multi-Spatial Attention Convolutional Neural Network for Prostate Cancer ISUP Grading

“…Using only slide-level weak labels, rather than manually drawn ROIs, Li et al [6] designed an attention-based multi-resolution multiple instance learning model, not only predicting slide-level grading scores, but also providing visualization of relevant regions using inherent attention maps. Zhang et al [19] employed a Bi-Attention adversarial network for PCa segmentation, combining attention with a generative adversarial network (GAN). By using channel and position attention simultaneously in one network, key features of PCa regions can be highlighted globally and locally, resulting in satisfying segmentation performance.…”

“…Specifically, the MCMS Attention module boosts feature representation in two aspects: the model learns a channel attention vector to assign weights to channels in the feature map by pooling from multiple attention branches with different reduction ratios; similarly, the model also learns a spatial attention matrix to focus on more relevant areas of the image, by pooling from multiple convolutional layers with different kernel sizes. Compared to other bi-attention-based models [19,23], our model mines feature information in a finer granularity and achieves superior performance.…”