Histopathological Synthetic Augmentation with Generative Models

Jiarong Ye (), Peng Jin (), Haomiao Ni (), Sharon X. Huang () and Yuan Xue ()
Additional contact information
Jiarong Ye: The Pennsylvania State University, College of Information Sciences and Technology
Peng Jin: The Pennsylvania State University, College of Information Sciences and Technology
Haomiao Ni: The University of Memphis, Department of Computer Science
Sharon X. Huang: The Pennsylvania State University, College of Information Sciences and Technology
Yuan Xue: The Ohio State University, Department of Biomedical Informatics

Chapter Chapter 10 in Generative Machine Learning Models in Medical Image Computing, 2025, pp 183-207 from Springer

Abstract: Abstract In this chapter, we explore the application of generative models to enhance the fidelity and utility of artificially generated images for data augmentation in digital pathology. We employ HistoGAN for synthetic augmentation, filtering images based on label congruence and feature resemblance to real specimens. This ensures that only the most accurate synthetic images are used. Additionally, we utilize reinforcement learning for automated quality checks, optimizing synthetic sample selection, and improving image classification outcomes. However, GANs have limitations, such as instability during training and the need for large annotated datasets for conditional generation. To address these issues, we transition to HistoDiffusion, a model that utilizes diffusion processes, which are more stable to train and reduce the risk of mode collapse common in GANs. Furthermore, unconditional diffusion models can be guided with smaller annotated datasets to enable conditional synthesis. HistoDiffusion generates more complex and nuanced images, enhancing realism and diversity. Through this exploration of generative AI techniques for synthetic augmentation, each model addresses the limitations of its predecessor, advancing the effectiveness of data augmentation in digital pathology.

Date: 2025
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:sprchp:978-3-031-80965-1_10

Ordering information: This item can be ordered from
http://www.springer.com/9783031809651

DOI: 10.1007/978-3-031-80965-1_10

Access Statistics for this chapter

More chapters in Springer Books from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().