Pathologist-level interpretable whole-slide cancer diagnosis with deep learning https://github.com/zizhaozhang/nmi-wsi-diagnosis
paper repository https://github.com/UCSD-AI4H/PathVQA
- Image source: 1,670 pathology images collected from two pathology textbooks: “Textbook of Pathology" and “Basic Pathology”, 3,328 pathology images collected from the PEIR7 digital library
- 32,799 question-answer pairs generated
- Avg. # of questions per image: 6.6
- Max # of questions for a single image: 14
- Min # of questions for a single image: 1
- Avg. # of words per question: 9.5
- Avg. # of words per answer: 2.5
- # of question types = 7
- Open-ended questions: what(40.9%), where(4%), when(0.9%), whose(0.6%), how(3%), how much/how many(0.9%).
- Close-ended questions: yes/no. The number of “yes" and “no" answers are balanced
- total # of images in downloaded PVQA (splitted): train: 3021, val: 992, test: 991. (6:2:2) Indexed (q2a, qid2a, qid2q, etc.)
paper repository https://github.com/masatsuneki/histopathology-image-caption
- Patches extracted from 991 Whole Slide Images (WSI) of H&E-stained gastric adenocarcinoma specimens with associated diagnostic captions extracted directly from existing medical reports.
- The slides were digitized into WSIs at a magnification of x20.
- All of collected cases were diagnosed as having adenocarcinoma and were reviewed by three pathologists to confirm the diagnoses.
- adenocarcinoma subtypes: 9
- The vocabulary consisted of 277 words with a maximum sentence length of 50 words.
- Patch size: 300*300
- Available dataset contains patches extracted from WSI at 20x. No 10x patches
- # of tiles: 262777
- Captions columns: wsi_id, subtype ,text
- Patch filename convention: {wsi_id}_{random_hash}
Not available
| title | date | #views | #downloads | data | link | paper link | Code repository |
|---|---|---|---|---|---|---|---|
| PESO: Prostate Epithelium Segmentation on H&E-stained prostatectomy whole slide images | 26-Jul-21 | 8060 | 21960 | WSI & masks | https://zenodo.org/record/5137717#.ZCTnTOxByDU | Epithelium segmentation using deep learning in H&E-stained prostate specimens with immunohistochemistry as reference standard | |
| The Single-Cell Pathology Landscape of Breast Cancer | 4-Nov-19 | 7047 | 51045 | ome-tiff &masks | https://zenodo.org/record/4607374#.ZCTnf-xByDU | https://github.com/BodenmillerGroup/SCPathology_publication | |
| Histological images for MSI vs. MSS classification in gastrointestinal cancer, FFPE samples | 7-Feb-19 | 10498 | 43835 | patch (wsi on https://portal.gdc.cancer.gov/) | https://zenodo.org/record/2530835#.ZCTm1uxByDU | ||
| Segmentation of Nuclei in Histopathology Images by deep regression of the distance map | 16-Feb-18 | 8108 | 4434 | patch(small slide) and mask(gt) | https://zenodo.org/record/2579118#.ZCTm3exByDU | ||
| 100,000 histological images of human colorectal cancer and healthy tissue | 7-Apr-18 | 29467 | 34556 | patch (&masks?) | https://zenodo.org/record/1214456#.ZCTmi-xByDU | ||
| Collection of textures in colorectal cancer histology | 26-May-16 | 14800 | 33858 | patch(small & large size) | https://zenodo.org/record/53169#.ZCTmuexByDU | ||
| BACH Dataset : Grand Challenge on Breast Cancer Histology images | 31-May-19 | 5829 | 8360 | patch(class labeled) wsi (region partially? Annotated) | https://zenodo.org/record/3632035#.ZCTy6-xByDU | ||
| Data for: Tang et al., Interpretable classification of Alzheimer's disease pathologies with a convolutional neural network pipeline | 1-Nov-18 | 2601 | 4782 | WSI & patch (class annotation) | https://zenodo.org/record/1470797#.ZCT2EuxByDU | Interpretable classification of Alzheimer’s disease pathologies with a convolutional neural network pipeline | https://github.com/keiserlab/plaquebox-paper |
| Digital Pathology Dataset for Prostate Cancer Diagnosis | 4-Feb-22 | 1022 | 620 | WSI & annotations & patch | https://zenodo.org/record/7152243#.ZCT3B-xByDU | An AI-assisted Tool For Efficient Prostate Cancer Diagnosis in Low-grade and Low-volume Cases |