Generative Adversarial Networks in Digital Pathology: A Survey on Trends and Future Potential

Abstract

Summary Image analysis in the field of digital pathology has recently gained increased popularity. The use of high-quality whole-slide scanners enables the fast acquisition of large amounts of image data, showing extensive context and microscopic detail at the same time. Simultaneously, novel machine-learning algorithms have boosted the performance of image analysis approaches. In this paper, we focus on a particularly powerful class of architectures, the so-called generative adversarial networks (GANs) applied to histological image data. Besides improving performance, GANs also enable previously intractable application scenarios in this field. However, GANs could exhibit a potential for introducing bias. Hereby, we summarize the recent state-of-the-art developments in a generalizing notation, present the main applications of GANs, and give an outlook of some chosen promising approaches and their possible future applications. In addition, we identify currently unavailable methods with potential for future applications.

Digital Image Analysis of the Proliferation Markers Ki67 and Phosphohistone H3 in Gastroenteropancreatic Neuroendocrine Neoplasms: Accuracy of Grading Compared With Routine Manual Hot Spot Evaluation of the Ki67 Index

“…Digital pathology has during recent years, been introduced at more pathology departments, and the numbers are increasing. 31 , 32 Manual counting is often difficult because of the high cellularity commonly encountered in these tumors. 30 One of the benefits of our DIA method of Ki67 is the automatic separation of stroma and stromal cells from tumor cells.…”

Intuitive Graphical Visualization of Transcriptomes by Nonlinear Dimensionality Reduction Exposes Relatedness between Human Placenta Tissues

“…The datasets we analyzed here provides useful resources of maternal-fetal interface study models for obstetricians and gynecologists (Biancotti Juan-Carlos et al 2010). Recently developed generative adversarial networks have made great breakthroughs in generating biomedical images (Tschuchnig et al 2020) and single cell sequencing data (Li et al 2020). As the acquisition of placental tissue, especially the chorionic villus in early pregnancy, is restricted by medical ethics, the relevant data sets from different laboratories comprehensively analyzed here provide a preliminary basis for the generation of relevant data sets with different genetic backgrounds and developmental stage (Yang et al 2020) based on generative adversarial networks in the future, which can effectively alleviate the difficulties of funding and ethical limitations in biological research.…”

Artificial Intelligence in Pathology: From Prototype to Product

“…[ 28 29 ] Another recent approach is using Generative Adversarial Networks to automatically create realistic synthetic histology images. [ 30 31 ]…”