pT1 Hotspot Tumor Budding T-cell Graph (pT1-HBTG) Dataset

Dataset described in MIDL 2023 publication "Tumor Budding T-cell Graphs: Assessing the Need for Resection in pT1 Colorectal Cancer Patients". For more information, please refer to the article. Please cite this article when using the data set.

The code for the Graph Neural Network experiments can be found on GitHub: https://github.com/digitalpathologybern/pT1-HBTG-MIDL2023.

Content:

Zip file containing the different graph configurations in GXL format.

Node labels:

Coordinates x and y in \(\mu m\) relative to the top-left of the hotspot (0/0).

Type: lymphocyte/tumor bud (based on automated detection)

ImageNet ViT-26 (DINO) features

Edge labels:

Distance between the nodes in \(\mu m\)

JSON file with the class labels and cross-validation splits. The first level contains the split (5-folds, index 0-4), the second level the class labels (0/1), and the list of corresponding file-IDs. 

{
"0": {
"0": [
"107_1",
"107_2",
...
],
"1": [
"160",
"178_1",
"178_2",
...
]
},

...

"4": {
"0": [
"10",
"103",
"104",
...
],
"1": [
"117",
"140_1",
"140_2",
...
]
}
}

Zip file containing the PNGs of the ITBCC hotspots on which the graphs are based, extracted at full resolution (level 0, area of 0.785 mm2). The WSIs were digitized using a Pannoramic 250 scanner at \(0.243\mu m/pixel\).

Zip file containing the patches from which the ImageNet-based features were extracted (size 200x200 pixels, centered on the coordinates of the element)

File-ID nomenclature
The first number indicates the patient/case (e.g. 10.gxl). If we have more than one WSI per patient, they are indicated by a second number (e.g. 13_1.gxl and 13_2.gxl). The file-ID is consistent between all data.

Bibtex for citation:

@inproceedings{studer2023tumor,
title={Tumor Budding T-cell Graphs: Assessing the Need for Resection in pT1 Colorectal Cancer Patients},
author={Studer, Linda and Bokhorst, John-Melle and Nagtegaal, Iris and Zlobec, Inti and Dawson, Heather and Fischer, Andreas},
booktitle={Medical Imaging with Deep Learning},
year={2023}
}