Improved HER2 Tumor Segmentation with Subtype Balancing Using Deep Generative Networks
- Resource Type
- Conference
- Authors
- Ottl, Mathias; Steenpass, Jana; Rubner, Matthias; Geppert, Carol I.; Qiu, Jingna; Wilm, Frauke; Hartmann, Arndt; Beckmann, Matthias W.; Fasching, Peter A.; Maier, Andreas; Erber, Ramona; Breininger, Katharina
- Source
- 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI) Biomedical Imaging (ISBI), 2023 IEEE 20th International Symposium on. :1-5 Apr, 2023
- Subject
- Bioengineering
Computing and Processing
Photonics and Electrooptics
Signal Processing and Analysis
Image segmentation
Data privacy
Histopathology
Biological system modeling
Training data
Generative adversarial networks
Data models
HER2
Subtypes
Generative Models
Diffusion Models
Segmentation
- Language
- ISSN
- 1945-8452
Tumor segmentation in histopathology images is often complicated by its composition of different histological subtypes and class imbalance. Oversampling subtypes with low prevalence features is not a satisfactory solution since it eventually leads to overfitting. We propose to create synthetic images with semantically-conditioned deep generative networks and to combine subtype-balanced synthetic images with the original dataset to achieve better segmentation performance. We show the suitability of Generative Adversarial Networks (GANs) and especially diffusion models to create realistic images based on subtype-conditioning for the use case of HER2-stained histopathology. Additionally, we show the capability of diffusion models to conditionally inpaint HER2 tumor areas with modified subtypes. Combining the original dataset with the same amount of diffusion-generated images increased the tumor Dice score from 0.833 to 0.854 and almost halved the variance between the HER2 subtype recalls. These results create the basis for more reliable automatic HER2 analysis with lower performance variance between individual HER2 subtypes.