ParamNet: A Dynamic Parameter Network for Fast Multi-to-One Stain Normalization

TL;DR

This work tackles color variability in digital pathology by introducing ParamNet, a dynamic-parameter network that enables fast, multi-domain stain normalization. It combines a parameter-predictor sub-network with a compact 1×1 color-mapping module, allowing automatic generation of normalization parameters per image and achieving substantial speed (e.g., normalizing a 100,000×100,000 WSI in ~25s). An adversarial training framework with texture modules and four loss terms stabilizes training and preserves content while aligning color style, demonstrated across four diverse datasets. ParamNet consistently improves target similarity, source-content preservation, and downstream classifier accuracy, while maintaining high computational efficiency, indicating strong practical potential for clinical CAD pipelines.

Abstract

In practice, digital pathology images are often affected by various factors, resulting in very large differences in color and brightness. Stain normalization can effectively reduce the differences in color and brightness of digital pathology images, thus improving the performance of computer-aided diagnostic systems. Conventional stain normalization methods rely on one or several reference images, but one or several images may not adequately represent the entire dataset. Although learning-based stain normalization methods are a general approach, they use complex deep networks, which not only greatly reduce computational efficiency, but also risk introducing artifacts. Some studies use specialized network structures to enhance computational efficiency and reliability, but these methods are difficult to apply to multi-to-one stain normalization due to insufficient network capacity. In this study, we introduced dynamic-parameter network and proposed a novel method for stain normalization, called ParamNet. ParamNet addresses the challenges of limited network capacity and computational efficiency by introducing dynamic parameters (weights and biases of convolutional layers) into the network design. By effectively leveraging these parameters, ParamNet achieves superior performance in stain normalization while maintaining computational efficiency. Results show ParamNet can normalize one whole slide image (WSI) of 100,000x100,000 within 25s. The code is available at: https://github.com/khtao/ParamNet.

Figures (7)

• Figure 1: Network structure of ParamNet. First, the parameter prediction sub-network predicts the parameters (weights and biases of convolutional layers) of the color mapping sub-network at low resolution. Then, the color mapping sub-network uses these parameters to normalize the input image at original resolution.
• Figure 2: Adversarial training framework of ParamNet. Images from source domain are mapped to the target domain, after which texture details are added, and the result is mapped back to the source domain. The same reverse process is also performed for the images from the target domain. The difference is that in the inverse process, the texture module uses $T_B$ instead of $T_A$, and the discriminator uses $D_B$ instead of $D_A$.
• Figure 3: Visual comparison among different methods on the aligned cytopathology dataset.
• Figure 4: Visual comparison among different methods on the aligned histopathology dataset.
• Figure 5: Visual comparison among different methods on the cytopathology classification dataset. The source images have multiple styles (D2-D5), and the target images are from D1. Red boxes mark obviously abnormal normalized results.