Skip to content

Latest commit

 

History

History
135 lines (97 loc) · 6.83 KB

ANHIR.md

File metadata and controls

135 lines (97 loc) · 6.83 KB
Raw

ANHIR

Dataset Information

The Automatic Non-rigid Histological Image Registration (ANHIR) dataset is a collection designed for registering histological images of lung lobes and breast tissue, and it was part of a challenge at ISBI 2019. The dataset contains 481 pairs of registration data, with images and labels provided for 222 training cases, while 259 validation cases come with images only.

In the field of digital pathology, one of the simplest yet most useful tasks is the visual comparison of consecutive tissue sections, which requires aligning all images to a common frame. Image alignment is also relevant for applications such as 3D reconstruction and image fusion. It enables pathologists to assess histology and expression of multiple biomarkers within the same region. Moreover, slices may undergo non-linear deformations due to tissue processing and pre-analytical steps, meaning they stretch and warp from one slice to another. Currently, only a few automated alignment tools are capable of handling large images with sufficient accuracy and reasonable processing time.

The ANHIR challenge focuses on comparing the accuracy and speed of automated non-linear registration methods on large images stained with different biomarkers from the same tissue sample. Manual landmarks annotated on the images will be used to evaluate the accuracy of the registrations. Robustness of the methods will be estimated by computing the number of times the final image alignment is improved by the performed registration operations. As a secondary criterion, the computation time will also be measured. The challenge requires all methods to operate fully automatically, without interaction or specific parameter settings for specific images (such as placing key points or adjusting parameters for some special images).

Dataset Meta Information

Dimensions Modality Task Type Anatomical Structures Anatomical Area Data Volume File Format
2D Histopathology Registration Cell, Tissue Chest 481 JPG, PNG

Resolution Details

Magnification Image size
Image Statistics 40 times 15k x 15k--50k x 50k Pixels

Detailed information on pathological images of different tissues:

Name Tissue Scanner Magnitude Resolution [µm/pixel] Avg. size [pixels]
lung-lesion_ Lung lesion Zeiss Axio Imager M1 40x 0.174 18k×15k
lung-lobes_ Whole mice lung lobes Zeiss Axio Imager M1 10x 1.274 11k×6k
mammary-glands_ Mammary glands Zeiss Axio Imager M1 10x 2.294 12k×4k
mice-kidney_ Mice kidney NanoZoomer 2.0HT 20x 0.227 37k×30k
COAD_ COlon ADenocarcinoma (colon cancer) 3DHistec Pannoramic 10x 0.468 60k×50k
gastric_ Gastric mucosa and gastric adenocarcinoma tissue fragments Leica Biosystems 40x 0.2528 60k×75k
breast_ Human breast Leica Biosystems 40x 0.2528 65k×60k
kidney_ Human kidney Leica Biosystems 40x 0.2528 18k×55k

Label Information Statistics

The data for the registration task mainly include the initial image, the image after registration, and the paired landmarks of the two images.

The landmarks mainly represent the correspondence between points in the two images but do not include all points. The format of the landmark file is as follows:

,X,Y
1,226,173
2,256,171
3,278,182
4,346,207
...

Visualization

Original Image.

Target Image.

File Structure

The challenge has prepared original size versions of 100%, 50%, 25%, 10%, and 5% reductions, with different versions corresponding to different scale-xpc. The folder includes both original and registration images.

DATASET
|- lesions_1
|   |- scale-5pc
|   |   |- 29-041-Izd2-w35-CD31-3-les1.jpg
|   |   |- 29-041-Izd2-w35-CD31-3-les1.csv
|   |   |- 29-041-Izd2-w35-CD31-3-les1.jpg
|   |   |- 29-041-Izd2-w35-CD31-3-les1.csv
|   |   | ...
|   |   |- 29-041-Izd2-w35-CD31-3-les1.jpg
|   |   '- 29-041-Izd2-w35-CD31-3-les1.csv
|   |- scale-10pc
|   | ...
|   '- scale-100pc
|   |- 29-041-Izd2-w35-CD31-3-les1.png
|   |- 29-041-Izd2-w35-CD31-3-les1.csv
|   | ...
|   |- 29-041-Izd2-w35-CD31-3-les1.png
|   '- 29-041-Izd2-w35-CD31-3-les1.csv
|- lesions_2
| ...
'- mammary-gland_2

Authors and Institutions

Jiří Borovec (Czech Technical University, Czech Republic)

Jan Kybic (Czech Technical University, Czech Republic)

Ignacio Arganda-Carreras (University of the Basque Country, Spain)

Dmitry V Sorokin (Lomonosov Moscow State University, Russia)

Source Information

Official Website: https://anhir.grand-challenge.org/

Download Link: https://anhir.grand-challenge.org/Download/

Article Address: https://ieeexplore.ieee.org/abstract/document/9058666

Publication Date: 2020-07

Citation

@article{borovec2020anhir,
  title={ANHIR: automatic non-rigid histological image registration challenge},
  author={Borovec, Ji{\v{r}}{\'\i} and Kybic, Jan and Arganda-Carreras, Ignacio and Sorokin, Dmitry V and Bueno, Gloria and Khvostikov, Alexander V and Bakas, Spyridon and Eric, I and Chang, Chao and Heldmann, Stefan and others},
  journal={IEEE transactions on medical imaging},
  volume={39},
  number={10},
  pages={3042--3052},
  year={2020},
  publisher={IEEE}
}

Original introduction article is here.