This repository contains lists of resources (including Datasets and Code Bases) that can help domain generalization research in computational pathology. These resources and their related concepts are further explained in the following manuscript:
@misc{jahanifar2023domain,
title={Domain Generalization in Computational Pathology: Survey and Guidelines},
author={Mostafa Jahanifar and Manahil Raza and Kesi Xu and Trinh Vuong and Rob Jewsbury and Adam Shephard and Neda Zamanitajeddin and Jin Tae Kwak and Shan E Ahmed Raza and Fayyaz Minhas and Nasir Rajpoot},
year={2023},
eprint={2310.19656},
archivePrefix={arXiv},
primaryClass={eess.IV}
}
Any contribution will be appreciated. To contribute to this awesome list or suggest new resources, please make a PR and add your suggestions.
Publicly available datasets for DG experiments in CPath. Column DS represents the type domain shift that can be studied with each dataset (1: Covariate Shift, 2: Prior Shift, 3: Posterior Shift, and 4: Class-Conditional Shift).
| Dataset | Application/Task | DS | Domains |
|---|---|---|---|
| Detection | |||
| ATYPIA14 [paper][download] | Mitosis detection in breast cancer | 1 | 2 scanners |
| Crowdsource [paper] | Nuclei detection in renal cell carcinoma | 3 | 6 annotators |
| TUPAC-Aux [paper][download] | Mitosis detection in breast cancer | 1 | 3 centers |
| DigestPath [paper][download] | Signet ring cell detection in colon cancer | 1 | 4 centers |
| TiGER-Cells [paper][download] | TILs detection in breast cancer | 1 | 3 sources |
| MIDOG [paper][download] | Mitosis detection in multiple cancer types | 1, 2, 3 | 7 tumors, 2 species |
| Classification | |||
| TUPAC-Mitosis [paper][download] | BC proliferation scoring based on mitosis score | 1 | 3 centers |
| Camelyon16 [paper][download] | Lymph node WSI classification for BC metastasis | 1 | 2 centers |
| PatchCamelyon [paper][download] | BC tumor classification based on Camelyon16 | 1 | 2 centers |
| Camelyon17 [paper][download] | BC metastasis detection and pN-stage estimation | 1 | 5 centers |
| LC25000 [paper][download] | Lung and colon tumor classification | 4 | 2 organs |
| Kather 100K [paper][download] | Colon cancer tissue phenotype classification | 1 | 3 centers |
| WILDS [paper][download] | BC tumor classification based on Camelyon17 | 1 | 5 centers |
| PANDA [paper][download] | ISUP and Gleason grading of prostate cancer | 1, 2, 3 | 2 centers |
| Regression | |||
| TUPAC-PAM50 [paper][download] | BC proliferation scoring based on PAM50 | 1 | 3 centers |
| LYSTO [paper][download] | Lymphocyte assessment (counting) in IHC images | 1 | 3 cancers, 9 centers |
| CoNIC (Lizard) [paper][download] | Cellular composition in colon cancer | 1, 3 | 6 sources |
| TiGER-TILs [paper][download] | TIL score estimation in breast cancer | 1 | 3 sources |
| Segmentation | |||
| Crowdsource [paper] | Nuclear segmentation in renal cell carcinoma | 3 | 6 annotators |
| Camelyon [paper][download] | BC metastasis segmentation in lymph node WSIs | 1 | 2 and 5 centers |
| DS Bowl 2018 [paper][download] | Nuclear instance segmentation | 1, 4 | 31 sets, 5 modalities |
| CPM [paper][download] | Nuclear instance segmentation | 1, 4 | 4 cancers |
| BCSS [paper][download] | Semantic tissue segmentation in BC (from TCGA) | 1 | 20 centers |
| AIDPATH [paper] | Glomeruli segmentation in Kidney biopsies | 1 | 3 centers |
| PanNuke [paper][download] | Nuclear instance segmentation and classification | 1, 2, 4 | 19 organs |
| MoNuSeg [paper][download] | Nuclear instance segmentation in H&E images | 1 | 9 organs, 18 centers |
| CryoNuSeg [paper][download] | Nuclear segmentation in cryosectioned H&E | 1, 3 | 10 organs, 3 annotations |
| MoNuSAC [paper][download] | Nuclear instance segmentation and classification | 1, 2 | 37 centers, 4 organs |
| Lizard [paper][download] | Nuclear instance segmentation and classification | 1, 3 | 6 sources |
| MetaHistoSeg [paper][download] | Multiple segmentation tasks in various cancers | 1 | 5 sources/tasks |
| PANDA [paper][download] | Tissue segmentation in prostate cancer | 1, 2 | 2 centers |
| TiGER-BCSS [paper][download] | Tissue segmentation in BC (BCSS extension) | 1 | 3 sources |
| DigestPath [paper][download] | Colon tissue segmentation | 1 | 4 centers |
| NuInsSeg [paper][download] | Nuclear instance segmentation pan-cancer/species | 1,4 | 31 organs, 2 species |
| Survival and gene expression prediction | |||
| TCGA [papers][download] | Pan-cancer survival and gene expression prediction | 1, 2, 4 | 33 cancers, 20 centers |
| CPTAC [papers][download][tool] | Pan-cancer survival and gene expression prediction | 1, 2 | 10 cancers, 11 centers |
| Reference | DG Method | Title |
|---|---|---|
| Pretraining | ||
| Yang et al. [paper][code] | Minimizing Contrastive Loss | CS-CO: A Hybrid Self-Supervised Visual Representation Learning Method for H&E-stained Histopathological Images |
| Li et al. [paper][code] | Minimizing Contrastive Loss | Lesion-Aware Contrastive Representation Learning For Histopathology Whole Slide Images Analysis |
| Galdran et al. [paper][code] | Unsupervised/Self-supervised learning | Test Time Transform Prediction for Open Set Histopathological Image Recognition |
| Bozorgtabar et al. [paper][code] | Unsupervised/Self-supervised learning | SOoD: Self-Supervised Out-of-Distribution Detection Under Domain Shift for Multi-Class Colorectal Cancer Tissue Types |
| Koohbanani et al. [paper][code] | Multiple Pretext Tasks | Self Path: Self Supervision for Classification of Histology Images with Limited Budget of Annotation |
| Abbet et al. [paper][code] | Unsupervised/Self-supervised learning | Self-rule to multi-adapt: Generalized multi-source feature learning using unsupervised domain adaptation for colorectal cancer tissue detection |
| Cho et al. [paper][code] | Unsupervised/Self-supervised learning | Cell Detection in Domain Shift Problem Using Pseudo-Cell-Position Heatmap |
| Chikontwe et al. [paper][code] | Unsupervised/Self-supervised learning | Weakly supervised segmentation on neural compressed histopathology with self-equivariant regularization |
| Tran et al. [paper][code] | Minimizing Contrastive Loss | S5CL: Unifying Fully-Supervised, Self-Supervised, and Semi-Supervised Learning Through Hierarchical Contrastive Learning |
| Sikaroudi et al. [paper][code] | Unsupervised/Self-supervised learning | Supervision and Source Domain Impact on Representation Learning: A Histopathology Case Study |
| Wang et al. [paper][code] | Unsupervised/Self-supervised learning | Transformer-based unsupervised contrastive learning for histopathological image classification |
| Kang et al. [paper][code] | Unsupervised/Self-supervised learning | Benchmarking Self-Supervised Learning on Diverse Pathology Datasets |
| Lazard et al. [paper][code] | Contrastive Learning | Giga-SSL: Self-Supervised Learning for Gigapixel Images |
| Vuong et al. [paper][code] | Contrastive Learning | IMPaSh: A Novel Domain-Shift Resistant Representation for Colorectal Cancer Tissue Classification |
| Chen et al. [paper][code] | Unsupervised/Self-supervised learning | Fast and scalable search of whole-slide images via self-supervised deep learning |
| Meta-Learning | ||
| Sikaroudi et al. [paper][code] | Meta-learning | Magnification Generalization For Histopathology Image Embedding |
| Yuan et al. [paper][code] | Meta-learning | MetaHistoSeg: A Python Framework for Meta Learning in Histopathology Image Segmentation |
| Domain Alignment | ||
| Sharma et al. [paper][code] | Mutual Information | MaNi: Maximizing Mutual Information for Nuclei Cross-Domain Unsupervised Segmentation |
| Boyd et al. [paper][code] | Generative Models | Region-guided CycleGANs for Stain Transfer in Whole Slide Images |
| Kather et al. [paper][code] | Stain Normalization | Deep learning can predict microsatellite instability directly from histology in gastrointestinal cancer |
| Zheng et al. [paper][code] | Stain Normalization | Adaptive color deconvolution for histological WSI normalization |
| Sebai et al. [paper][code] | Stain Normalization | MaskMitosis: a deep learning framework for fully supervised, weakly supervised, and unsupervised mitosis detection in histopathology images |
| Zhang et al. [paper][code] | Minimizing Contrastive Loss | Stain Based Contrastive Co-training for Histopathological Image Analysis |
| Shahban et al. [paper][code] | Generative Models | Staingan: Stain Style Transfer for Digital Histological Images |
| Wagner et al. [paper][code] | Generative Models | Federated Stain Normalization for Computational Pathology |
| Quiros et al. [paper][code] | Domain Adversarial Learning | Adversarial learning of cancer tissue representations |
| Salehi et al. [paper][code] | Minimizing the KL Divergence | Unsupervised Cross-Domain Feature Extraction for Single Blood Cell Image Classification |
| Wilm et al. [paper][code] | Domain-Adversarial Learning | Domain adversarial retinanet as a reference algorithm for the mitosis domain generalization (midog) challenge |
| Haan et al. [paper][code] | Generative models | Deep learning-based transformation of H&E stained tissues into special stains |
| Dawood et al. [paper][code] | Stain Normalization | Do Tissue Source Sites leave identifiable Signatures in Whole Slide Images beyond staining? |
| Data Augmentation | ||
| Pohjonen et al. [paper][code] | Data augmentation | Augment like there’s no tomorrow: Consistently performing neural networks for medical imaging |
| Chang et al. [paper][code] | Stain Augmentation | Stain Mix-up: Unsupervised Domain Generalization for Histopathology Images |
| Shen et al. [paper][code] | Stain Augmentation | RandStainNA: Learning Stain-Agnostic Features from Histology Slides by Bridging Stain Augmentation and Normalization |
| Koohbanani et al. [paper][code] | Data augmentation | NuClick: A deep learning framework for interactive segmentation of microscopic images |
| Wang et al. [paper][code] | Data augmentation | A generalizable and robust deep learning algorithm for mitosis detection in multicenter breast histopathological images |
| Lin et al. [paper][code] | Generative Models | InsMix: Towards Realistic Generative Data Augmentation for Nuclei Instance Segmentation |
| Zhang et al. [paper][code] | Data augmentation | Benchmarking the Robustness of Deep Neural Networks to Common Corruptions in Digital Pathology |
| Yamashita et al. [paper][code] | Style Transfer Models | Learning domain-agnostic visual representation for computational pathology using medically-irrelevant style transfer augmentation |
| Falahkheirkhah et al. [paper][code] | Generative Models | Deepfake Histologic Images for Enhancing Digital Pathology |
| Scalbert et al. [paper][code] | Generative Models | Test-time image-to-image translation ensembling improves out-of-distribution generalization in histopathology |
| Mahmood et al. [paper][code] | Generative Models | Deep Adversarial Training for Multi-Organ Nuclei Segmentation in Histopathology Images |
| Fan et al. [paper][code] | Generative Models | Fast FF-to-FFPE Whole Slide Image Translation via Laplacian Pyramid and Contrastive Learning |
| Marini et al. [paper][code] | Stain Augmentation | Data-driven color augmentation for H&E stained images in computational pathology |
| Faryna et al. [paper][code] | RandAugment for Histology | Tailoring automated data augmentation to H&E-stained histopathology |
| Model Design | ||
| Graham et al. [paper][code] | Model design | Dense Steerable Filter CNNs for Exploiting Rotational Symmetry in Histology Images |
| Lafarge et al. [paper][code] | Model design | Roto-translation equivariant convolutional networks: Application to histopathology image analysis |
| Zhang et al. [paper][code] | Model design | DDTNet: A dense dual-task network for tumor-infiltrating lymphocyte detection and segmentation in histopathological images of breast cancer |
| Graham et al. [paper][code] | Model Design | One model is all you need: Multi-task learning enables simultaneous histology image segmentation and classification |
| Yu et al. [paper][code] | Model Design | Prototypical multiple instance learning for predicting lymph node metastasis of breast cancer from whole-slide pathological images |
| Yaar et al. [paper][code] | Model Design | Cross-Domain Knowledge Transfer for Prediction of Chemosensitivity in Ovarian Cancer Patients |
| Tang et al. [paper][code] | Model Design | Probeable DARTS with Application to Computational Pathology |
| Vuong et al. [paper][code] | Model Design | Joint categorical and ordinal learning for cancer grading in pathology images |
| Domain Separation | ||
| Wagner et al. [paper][code] | Generative Models | HistAuGAN: Structure-Preserving Multi-Domain Stain Color Augmentation using Style-Transfer with Disentangled Representations |
| Chikontwe et al. [paper][code] | Learning disentangled representations | Feature Re-calibration based Multiple Instance Learning for Whole Slide Image Classification |
| Ensemble Learning | ||
| Sohail et al. [paper][code] | Ensemble learning | Mitotic nuclei analysis in breast cancer histopathology images using deep ensemble classifier |
| Regularization Strategies | ||
| Mehrtens et al. [paper][code] | Regularization Strategies | Benchmarking common uncertainty estimation methods with histopathological images under domain shift and label noise |
| Other | ||
| Lu et al. [paper][code] | Other | Federated learning for computational pathology on gigapixel whole slide images |
| Aubreville et al. [paper][code] | Other | Quantifying the Scanner-Induced Domain Gap in Mitosis Detection |
| Sadafi et al. [paper][code] | Other | A Continual Learning Approach for Cross-Domain White Blood Cell Classification |