Datasets

Our research group created high-quantity, high-quality data sets of microscopy images.

A large-scale dataset for mitotic figure assessment on whole slide images of canine cutaneous mast cell tumor

A dataset comprising >44k mitotic figure annotations and >240k overall cell annotations. We’ve shown that an F1-score of 0.82 can be achieved on this data set using state-of-the-art deep learning methods.

Paper
Sci Data 6:274, 2019
data on figshare
code on github

A completely annotated whole slide image dataset of canine breast cancer to aid human breast cancer research

This whole slide image dataset comprises around 14k mitotic figure annotations, split across 21 cases of canine mammary carcinoma. We’ve shown in the paper that networks trained on this dataset may generalize to human tissue.

Paper
Sci Data 7:417, 2020
Dataset
data on figshare

Code
code on github

Dataset on Bi- and Multi-Nucleated Tumor Cells in Canine Cutaneous Mast Cell Tumors

This dataset extends our Large-Scale Canine Cutaneous Mast Cell Tumor Dataset by also incorporating full annotations about bi- and multinucleated cells, which play also a role in tumor prognostication.

Paper
BVM 2021 paper
Dataset
dataset and code on github

Computer-augmented label set for the TUPAC16 mitotic figure data set

We’ve created an alternative, high-quality label set using our well-established mitotic figure annotation workflow for the popular TUPAC16 mitotic figure data set (human mammary carcinoma).

Paper
LABELS@MICCAI 2020 paper
Dataset
Data and code on github