Table 2.
Examples of grand challenges held in CPATH.
Challenge | Year | Staining | Challenge Goal | Dataset |
---|---|---|---|---|
GlaS challenge [55] | 2015 | H&E | Segmentation of colon glands of stage T3 and T4 colorectal adenocarcinoma | Private set—165 images from 16 WSIs |
CAMELYON16 [56] | 2016 | H&E | Evaluation of new and current algorithms for automatic identification of metastases in WSIs from H&E-stained lymph node sections | Private set—221 images |
TUPAC challenge [57] | 2016 | H&E | Prediction of tumor proliferation scores and gene expression of breast cancer using histopathology WSIs | 821 TCGA WSIs |
BreastPathQ [58] | 2018 | H&E | Development of quantitative biomarkers to determinate cancer cellularity of breast cancer from H&E-stained WSIs | Private set—96 WSIs |
BACH challenge [59] | 2018 | H&E | Classification of H&E-stained breast histopathology images and performing pixel-wise labeling of WSIs | Private set—40 WSIs and 500 images |
LYON19 [60] | 2019 | IHC | Provision of a dataset as well as an evolution platform for current lymphocyte detection algorithms in IHC-stained images | LYON19 test set containing 441 ROIs |
DigestPath [61] | 2019 | H&E | Evaluation of algorithms for detecting signet ring cells and screening colonoscopy tissue from histopathology images of the digestive system | Private set—127 WSIs |
HEROHE ECDP [62] | 2020 | H&E | Evaluation of algorithms to discriminate HER2-positive breast cancer specimens from HER2-negative breast cancer specimens with high sensitivity and specificity only using H&E-stained slides | Private set—359 WSIs |
MIDOG challenge [63] | 2021 | H&E | Detection of mitotic figures from breast cancer histopathology images scanned by different scanners to overcome the ‘domain-shift’ problem and improve generalization | Private set—200 cases |
CoNIC challenge [64] | 2022 | H&E | Evaluation of algorithms for nuclear segmentation and classification into six types, along with cellular composition prediction | 4981 patches |
ACROBAT [65] | 2022 | H&E, IHC | Development of WSI registration algorithms that can align WSIs of IHC-stained breast cancer tissue sections with corresponding H&E-stained tissue regions | Private dataset—750 cases consisting of 1 H&E and 1–4 matched IHC |