Overview
SegPath is created for the semantic segmentation of H&E images for eight major cell types in tumor tissue.
The dataset is constructed by immunofluorescence restaing. First, sections were stained with H&E. They were then digitized using a slide scanner to create whole slide images (WSIs). After destaining the H&E-stained sections with alcohol and autoclave processing, IF and 4',6-diamidino-2-phenylindole dihydrochloride (DAPI) nuclear staining were performed using antibodies that specifically recognized each cell type. The slides were then digitized again. Multiresolution rigid registration between the H&E and IF images was performed to ensure that the haematoxylin component in the H&E images and DAPI in the IF images, both recognizing nuclei, had been aligned.
Cell types in SegPath
epithelial cells
(anti-panCK)
smooth muscle cells/myofibroblasts
(anti-αSMA)
red blood cells
(anti-CD235a)
leukocytes
(anti-CD45RB)
lymphocytes
(anti-CD3/CD20)
endothelial cells
(anti-ERG)
plasma cells
(anti-MIST1)
myeloid cells
(anti-MNDA)
epithelial cells (anti-panCK)
DL link :
https://zenodo.org/record/7412731
Dataset organization
A Tar.gz file contains the following files:
- HE image file:
{antigen}_{celltype}_{slideID}_{posx}_{posy}_HE.png
- Mask image file:
{antigen}_{celltype}_{slideID}_{posx}_{posy}_mask.png
Each image file is 984x984 px.
posX and posY are the leftmost position in WSI coordinate.
Mask files store binary segmentation mask (background : 0,
target : 1)
A csv file contains the following information:
antigen : Antibodies for this antigen were used to create the
segmentation mask.
filename: filename of image or mask file.
train_val_test : train, validation, or test sample in the paper.
Licenses
This work is licensed under a Creative Commons
Attribution-NonCommercial-ShareAlike 4.0 International (CC-BY-NC-SA 4.0)
For non-commercial use, please use the dataset under
CC-BY-NC-SA.
If you would like to use the dataset for commercial purposes,
please contact us (ishum-prm@m.u-tokyo.ac.jp).