{"success":true,"data":{"author":"Le Hou, Joel Saltz, Tahsin Kurc et al. (Stony Brook University); mirrored by MedOtter","bucketAccessType":"public","bucketName":"MedOtter/Pan-Cancer-Nuclei-Seg","bucketPrefix":"","category":"Health & Biomedical","cloudProvider":"huggingface","createdAt":"2026-11-21T20:45:56.170Z","description":"The manually segmented portion of the Pan-Cancer-Nuclei-Seg resource from Stony Brook Biomedical Informatics, published through The Cancer Imaging Archive (TCIA). It has 1,356 H&E patches of 256x256 pixels (40x, about 0.25 microns per pixel) cropped from TCGA whole-slide images, each with a consensus per-nucleus instance mask corrected by hand from Mask R-CNN output. Fourteen cancer types are balanced at roughly 91 to 100 patches each: BLCA, BRCA, CESC, COAD, GBM, LUAD, LUSC, PAAD, PRAD, READ, SKCM, STAD, UCEC and UVM. Each row records the TCGA slide barcode and crop coordinates, so patches can be traced to their source slide and deduplicated against other TCGA nucleus sets; 27 patches also have per-annotator masks at the source and 32 contain no nuclei. The algorithm-generated whole-slide segmentations of about 5 billion nuclei are not included. Parquet under data/ with 16-bit instance masks, about 160 MB (measured 2026-11-21); read masks with numpy rather than converting to 8-bit. Use for pan-cancer nucleus instance segmentation benchmarks. TCIA lists the collection under CC BY 3.0 with DOI 10.7937/TCIA.2019.4A4DKP9U; cite Hou et al., Scientific Data 2020 (doi 10.1038/s41597-020-0528-1).","displayTitle":"Pan-Cancer-Nuclei-Seg manual subset: 1,356 H&E patches with hand-corrected nucleus masks across 14 TCGA cancer types (TCIA)","doi":"10.7937/TCIA.2019.4A4DKP9U","huggingFaceUrl":"https://huggingface.co/datasets/MedOtter/Pan-Cancer-Nuclei-Seg","id":"2b106569-409a-53ae-8d6b-a4def91b3a0e","license":"CC-BY-3.0","licenseEvidenceUrl":"https://www.cancerimagingarchive.net/analysis-result/pan-cancer-nuclei-seg/","objectCountEstimate":3,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1038/s41597-020-0528-1","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/Pan-Cancer-Nuclei-Seg\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"MedOtter/Pan-Cancer-Nuclei-Seg\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=MedOtter, license:cc-by-3.0, size_categories:1K<n<10K, format:parquet, downloads=170"],"format":"mixed","generatedAt":"2026-11-21T20:48:39.603Z","layout":{"example":"data/train-00000-of-00001.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/train-00000-of-00001.parquet","truncatedAt":512}],"sampleObjects":["data/train-00000-of-00001.parquet"],"sidecars":[{"bytes":"---\nlicense: cc-by-3.0\ntask_categories:\n  - image-segmentation\ntags:\n  - medical\n  - histopathology\n  - h-and-e\n  - nuclei\n  - nucleus-segmentation\n  - instance-segmentation\n  - pan-cancer\n  - tcga\npretty_name: Pan-Cancer-Nuclei-Seg (manual subset)\nsize_categories:\n  - 1K<n<10K\n---\n\n# Pan-Cancer-Nuclei-Seg — Manual Subset\n\nGold-standard **manual** nucleus segmentation patches from the TCIA\n*Pan-Cancer-Nuclei-Seg* resource (Hou et al., *Scientific Data* 2020): **1,356**\nmanually segmented **256×256 H&E** patches with per-nucleus instance masks,\nspanning **14 TCGA cancer types**.\n\n> **Scope — please read.** This repository contains **only the manual subset**,\n> not the full Pan-Cancer-Nuclei-Seg resource. The headline \"~5 billion nuclei\"\n> component (5,060 whole-slide images, ~666 GB of CSV polygon vertices) is\n> **algorithm-generated** (a U-Net pipeline, *not* manual ground truth), is\n> distributed **without source images** (the H&E WSIs live in TCGA/GDC), and is\n> not hosted here. The 1,356 manual patches are the only part that is both\n> gold-standard and image-paired.\n\n- **Modality:** Histopathology — H&E brightfield, 256×256 patches at 40× (~0.25 µm/px)\n- **Target:** nucleus segmentation (single foreground class)\n- **Ground truth:** manually corrected Mask R-CNN masks agreed by annotators A, B\n  and C collectively (per the dataset readme)\n- **License:** CC BY 3.0\n- **Source:** [TCIA Pan-Cancer-Nuclei-Seg](https://www.cancerimagingarchive.net/analysis-result/pan-cancer-nuclei-seg/)\n  · Stony Brook BMI [Box `cnn-nuclear-segmentations-2019`](https://stonybrookmedicine.app.box.com/v/cnn-nuclear-segmentations-2019)\n\n## Cancer types (14)\nBLCA, BRCA, CESC, COAD, GBM, LUAD, LUSC, PAAD, PRAD, READ, SKCM, STAD, UCEC, UVM\n(~91–100 patches each, balanced).\n\n## Columns\n| Column | Type | Notes |\n|---|---|---|\n| `patch_id` | int32 | Patch-ID from the dataset readme (non-contiguous, 1..1365 with gaps) |\n| `image` | Image (RGB) | 256×256 H&E patch (`{id}_crop.png`), original PNG bytes |\n| `instance_mask` | Image (16-bit, mode `I;16`) | Consensus manual mask (`{id}_labeled_mask_corrected.png`); 0 = background, 1..N = per-nucleus instance IDs. Original PNG bytes (lossless) |\n| `cancer_type` | string | One of the 14 TCGA codes (lowercase) |\n| `wsi_id` | string | **TCGA slide barcode** the patch was cropped from (e.g. `TCGA-2F-A9KR-01Z-00-DX1`) — use as the cross-dataset dedup key |\n| `x`, `y` | int32 | Top-left coordinate of the crop in the source WSI |\n| `size_original` | int32 | Crop side length in source-WSI pixels before resize |\n| `size_in_40x` | int32 | Crop side length at 40× (400) |\n| `num_instances` | int32 | Number of nuclei in the patch (`= instance_mask.max()`) |\n| `has_multirater` | bool | True for the 27 patches that also have per-annotator masks at the source |\n\nFor binary nucleus-vs-background segmentation, treat `instance_mask > 0` as\nforeground. **Read the mask via numpy** (`np.array(mask)`) — passing a 16-bit\n`I;16` PNG through `PIL.Image.convert(\"L\")` divides values by 256 and erases most\ninstance IDs.\n\n**Empty masks:** 32 of the 1,356 patches contain no annotated nuclei\n(`num_instances == 0`, background/stroma-only tissue) — these are faithful to the\nsource, not a packaging error. Filter on `num_instances > 0` if your task\nrequires at least one nucleus (instance counts range 0–141, mean ≈ 29).\n\n## Provenance, naming and cross-dataset overlap\n- **Provenance:** official, author-hosted (Saltz/Kurc, Stony Brook BMI). 1,356\n  patches matches the paper.\n- **Faithful naming:** this is the *manual subset only* (see scope note above).\n- **Ground-truth tier:** the bulk Pan-Cancer-Nuclei-Seg masks are\n  algorithm-generated; **only these 1,356 patches are manual**. 27 patches carry\n  additional per-annotator masks at the source.\n- **Overlap (leakage hazard):** every patch is **TCGA-derived**, so it shares\n  source slides with other TCGA histopathology sets — notably **PanNuke**,\n  **MoNuSeg**, **MoNuSAC**, **NuCLS** and **TIGER** (on BRCA/PRAD/BLCA/LUAD).\n  Deduplicate against those by matching the **`w","encoding":"utf8","key":"README.md","name":"README.md","size":4911,"truncated":true}],"sourceObjectKey":"data/train-00000-of-00001.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/Pan-Cancer-Nuclei-Seg\")\nprint(ds)","starterSamplePath":"data/train-00000-of-00001.parquet","totalObjects":3},"summary":"The manually segmented portion of the Pan-Cancer-Nuclei-Seg resource from Stony Brook Biomedical Informatics, published through The Cancer Imaging Archive (TCIA).","title":"Pan-Cancer-Nuclei-Seg manual subset: 1,356 H&E patches with hand-corrected nucleus masks across 14 TCGA cancer types (TCIA)","updatedAt":"2026-11-22T01:08:39.895Z","visibility":"public","websiteUrl":"https://www.cancerimagingarchive.net/analysis-result/pan-cancer-nuclei-seg/"}}