AUTHOR=Jiang Liwen , Huang Shuting , Luo Chaofan , Zhang Jiangyu , Chen Wenjing , Liu Zhenyu 

TITLE=An improved multi-scale gradient generative adversarial network for enhancing classification of colorectal cancer histological images

JOURNAL=Frontiers in Oncology

VOLUME=Volume 13 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1240645

DOI=10.3389/fonc.2023.1240645

ISSN=2234-943X

ABSTRACT=In recent years, deep learning-based solutions for histological image classification have gained attention as they enable the objective evaluation of histological images. However, deep learning-based automatic histological image classification requires a large number of expert annotations, which can be both time-consuming and labor-intensive to obtain. Currently, several scholars have proposed generative models to augment labeled data. However, images synthesized by generators may not learn the data distribution completely, resulting in label uncertainty, which means the generated images contain different class information and cannot be categorized as any particular class. To alleviate these issues, a method called InceptionV3-SMSG-GAN has been proposed to enhance classification performance by obtaining generated images with higher quality. Specifically, images synthesized by Multi-Scale Gradients Generative Adversarial Network (MSG-GAN) are selectively added to the training set through a selection mechanism utilizing a trained model to choose generated images with higher class probabilities. The selection mechanism filters the synthetic images that contain ambiguous category information, thus alleviating label uncertainty. Experimental results show that compared with the baseline method which uses InceptionV3, the proposed method can significantly improve the performance of pathological image classification from 86.87% to 89.54% for overall accuracy. Additionally, the quality of generated images is evaluated quantitatively using various commonly used evaluation metrics.