Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DeeperHistReg: Robust Whole Slide Images Registration Framework

Marek Wodzinski, Niccolò Marini, Manfredo Atzori, Henning Müller

TL;DR

Problem: robust automatic registration of WSIs across different stains remains challenging due to large size and cross-stain variability. Approach: DeeperHistReg provides a configurable pipeline combining preprocessing, initial affine alignment, and nonrigid registration, capable of handling WSIs up to $200k × 200k$ and saving results as pyramid TIFFs. Contributions: an extensible, ready-to-use library with OpenSlide/PyVips IO, JSON-configured workflows, and PyPI/Docker availability, incorporating state-of-the-art ANHIR/ACROBAT methods and support for annotation transfer and segmentation. Impact: enables annotation transfer, 3-D reconstruction, content-based retrieval, and improved multimodal AI in digital pathology; the authors report top challenge performance and plan future BioFormats support and a dedicated 3-D reconstruction library.

Abstract

DeeperHistReg is a software framework dedicated to registering whole slide images (WSIs) acquired using multiple stains. It allows one to perform the preprocessing, initial alignment, and nonrigid registration of WSIs acquired using multiple stains (e.g. hematoxylin \& eosin, immunochemistry). The framework implements several state-of-the-art registration algorithms and provides an interface to operate on arbitrary resolution of the WSIs (up to 200k x 200k). The framework is extensible and new algorithms can be easily integrated by other researchers. The framework is available both as a PyPI package and as a Docker container.

DeeperHistReg: Robust Whole Slide Images Registration Framework

TL;DR

Problem: robust automatic registration of WSIs across different stains remains challenging due to large size and cross-stain variability. Approach: DeeperHistReg provides a configurable pipeline combining preprocessing, initial affine alignment, and nonrigid registration, capable of handling WSIs up to and saving results as pyramid TIFFs. Contributions: an extensible, ready-to-use library with OpenSlide/PyVips IO, JSON-configured workflows, and PyPI/Docker availability, incorporating state-of-the-art ANHIR/ACROBAT methods and support for annotation transfer and segmentation. Impact: enables annotation transfer, 3-D reconstruction, content-based retrieval, and improved multimodal AI in digital pathology; the authors report top challenge performance and plan future BioFormats support and a dedicated 3-D reconstruction library.

Abstract

DeeperHistReg is a software framework dedicated to registering whole slide images (WSIs) acquired using multiple stains. It allows one to perform the preprocessing, initial alignment, and nonrigid registration of WSIs acquired using multiple stains (e.g. hematoxylin \& eosin, immunochemistry). The framework implements several state-of-the-art registration algorithms and provides an interface to operate on arbitrary resolution of the WSIs (up to 200k x 200k). The framework is extensible and new algorithms can be easily integrated by other researchers. The framework is available both as a PyPI package and as a Docker container.
Paper Structure (11 sections, 5 figures)

This paper contains 11 sections, 5 figures.

Table of Contents

  1. Motivation and significance
  2. Software description
  3. Overview & Software capabilities
  4. Software architecture
  5. Availability & Usage
  6. Illustrative examples
  7. Preprocessing
  8. Affine Registration
  9. Nonrigid Registration
  10. Computational details
  11. Impact and conclusions

Figures (5)

  • Figure 1: Overview of the WSI registration. The goal is to align the input images to the same geometrical space.
  • Figure 2: The DeeperHistReg processing pipeline.
  • Figure 3: Visualization of WSIs preproccesing using slides from the ACROBAT dataset. The images are initially resampled, padded and converted to grayscale.
  • Figure 4: Exemplary WSI initial alignment result using images from the ANHIR dataset. The images are initially aligned to the same geometrical space, to be further improved by the nonrigid registration.
  • Figure 5: Exemplary results of the deformable registration at various magnification levels using a re-stained H&E and PHH3 slides from the HyReCo database. Please note the spatial agreement even at the cell level.