CIMA histology images

Dataset Overview

Dataset Name

Dataset: histology landmarks

Dataset Content

The dataset contains 2D tissue micro‑section images stained by various methods. It is mainly used to assess registration quality and includes 108 images with manually placed landmarks.

Image Characteristics

Image Size: Extremely large
Visual Heterogeneity: Significant
Lack of Salient Objects: Yes

Landmark Information

Structure: Follows ImageJ standard structure and coordinate framework
Origin: Upper‑left corner of the image plane [0, 0]
Tools: Simple macros are provided for importing and exporting landmarks
File Layout: Each image has a corresponding .csv file stored in the same directory

Dataset Structure

[set_name]

Landmark Generation and Visualization

Generation: Use python handlers/run_generate_landmarks.py script
Visualization: Use python handlers/run_visualise_landmarks.py script

Annotation Details

Initial Annotations: Aggregated landmark sets placed by multiple users
Additional Annotations: Aimed at improving precision by comparing landmarks across differently stained images

Annotation Structure

[set_name]

Validation Process

Visual Check: Preliminary validation by comparing deformation of landmark pairs
Error Analysis: Compute inter‑landmark error; pairs exceeding a threshold are flagged as suspicious

References

J. Borovec, A. Muñoz‑Barrutia, and J. Kybic, “Benchmarking of image registration methods for differently stained histological slides” in 2018 25th IEEE International Conference on Image Processing (ICIP), p. 3368‑3372, 2018. DOI: 10.1109/ICIP.2018.8451040

Description