Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Auxiliary CycleGAN-guidance for Task-Aware Domain Translation from Duplex to Monoplex IHC Images

Nicolas Brieu, Nicolas Triltsch, Philipp Wortmann, Dominik Winter, Shashank Saran, Marlon Rebelatto, Günter Schmidt

TL;DR

This work proposes - through the introduction of a novel training design, an alternative constrain leveraging a set of immunofluorescence images as an auxiliary unpaired image domain for translation between images stained with chromogenic monoplex and duplex immunohistochemistry assays.

Abstract

Generative models enable the translation from a source image domain where readily trained models are available to a target domain unseen during training. While Cycle Generative Adversarial Networks (GANs) are well established, the associated cycle consistency constrain relies on that an invertible mapping exists between the two domains. This is, however, not the case for the translation between images stained with chromogenic monoplex and duplex immunohistochemistry (IHC) assays. Focusing on the translation from the latter to the first, we propose - through the introduction of a novel training design, an alternative constrain leveraging a set of immunofluorescence (IF) images as an auxiliary unpaired image domain. Quantitative and qualitative results on a downstream segmentation task show the benefit of the proposed method in comparison to baseline approaches.

Auxiliary CycleGAN-guidance for Task-Aware Domain Translation from Duplex to Monoplex IHC Images

TL;DR

This work proposes - through the introduction of a novel training design, an alternative constrain leveraging a set of immunofluorescence images as an auxiliary unpaired image domain for translation between images stained with chromogenic monoplex and duplex immunohistochemistry assays.

Abstract

Generative models enable the translation from a source image domain where readily trained models are available to a target domain unseen during training. While Cycle Generative Adversarial Networks (GANs) are well established, the associated cycle consistency constrain relies on that an invertible mapping exists between the two domains. This is, however, not the case for the translation between images stained with chromogenic monoplex and duplex immunohistochemistry (IHC) assays. Focusing on the translation from the latter to the first, we propose - through the introduction of a novel training design, an alternative constrain leveraging a set of immunofluorescence (IF) images as an auxiliary unpaired image domain. Quantitative and qualitative results on a downstream segmentation task show the benefit of the proposed method in comparison to baseline approaches.
Paper Structure (6 sections, 7 equations, 3 figures, 1 table)

This paper contains 6 sections, 7 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Results
  4. Conclusion
  5. Acknowledgments
  6. Compliance with ethical standards

Figures (3)

  • Figure 1: Auxiliary CycleGAN $\mathcal{G}_{AC}$ and $\mathcal{G}_{CA}$ between duplex IHC (A) and IF (C) image domain for the task-aware guidance of the GAN-based translation $\mathcal{G}_{AB}$ of duplex IHC images to the monoplex IHC image domain (B).
  • Figure 2: Qualitative results on representative duplex IHC images $x_A$, displayed at the foremost left column of each row. The synthetic monoplex IHC images $\mathcal{G}_{AB}(x_A)$ and the corresponding posterior maps $S_B\circ\mathcal{G}_{AB}(x_A)$ associated to the downstream nucleus segmentation task are shown for each baseline method $\textbf{b1-3}$. Predicted centers appear as holes in the posterior maps. For each method $\textbf{p1-3}$ the synthetic images $\mathcal{G}_{AC}(x_A)$ generated by the auxiliary CycleGAN are additionally shown first.
  • Figure 3: Left and middle: Sensitivity and specificity curves for different threshold values applied on the predicted nucleus posterior. Left: corresponding Receiver Operating Characteristic (ROC) curves.