{"success":true,"data":{"author":"Mohamed Amgad, Lee Cooper et al. (Northwestern University); mirrored by MedOtter","bucketAccessType":"public","bucketName":"MedOtter/NuCLS","bucketPrefix":"","category":"Health & Biomedical","cloudProvider":"huggingface","createdAt":"2026-11-21T20:45:56.144Z","description":"NuCLS (Nucleus Classification, Localization and Segmentation) is a crowdsourced and pathologist-reviewed annotation set of cell nuclei in breast cancer histology. This repository holds the corrected single-rater subset: 1,744 H&E fields of view cropped at 40x (about 0.2 microns per pixel) from 124 TCGA-BRCA whole-slide images, with 59,485 annotated nuclei. Nuclei are labeled in a 13-class scheme that collapses to 7 or 4 super-classes (tumor, fibroblast, lymphocyte, plasma cell, macrophage, mitotic figure, vascular endothelium, myoepithelium, apoptotic body, neutrophil, ductal epithelium, eosinophil). About one third of nuclei have true polygon outlines and two thirds are bounding boxes rasterized as filled rectangles; the per-FOV annotation CSV keeps the type so the two can be told apart. Masks are 3-channel PNGs encoding class, region of interest and instance id. Parquet under data/, about 400 MB (measured 2026-11-21); read with Hugging Face datasets or pandas plus Pillow. The uncorrected and multi-rater tiers remain at the official site. Use for nucleus detection, classification and tumor-infiltrating lymphocyte scoring. The authors release NuCLS under CC0 1.0; cite Amgad et al., GigaScience 2022 (doi 10.1093/gigascience/giac037).","displayTitle":"NuCLS: Nucleus Classification, Localization and Segmentation in TCGA breast cancer H&E (corrected single-rater set, 59k nuclei)","huggingFaceUrl":"https://huggingface.co/datasets/MedOtter/NuCLS","id":"827d0917-1786-5f6c-9714-89f51c81dd99","license":"CC0-1.0","licenseEvidenceUrl":"https://sites.google.com/view/nucls","objectCountEstimate":5,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1093/gigascience/giac037","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/NuCLS\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"MedOtter/NuCLS\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=MedOtter, license:cc0-1.0, size_categories:1K<n<10K, downloads=556"],"format":"mixed","generatedAt":"2026-11-21T20:48:38.918Z","layout":{"example":"data/train-00000-of-00003.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/train-00000-of-00003.parquet","truncatedAt":512}],"sampleObjects":["data/train-00000-of-00003.parquet","data/train-00001-of-00003.parquet","data/train-00002-of-00003.parquet"],"sidecars":[{"bytes":"---\nlicense: cc0-1.0\ntask_categories:\n  - image-segmentation\ntags:\n  - medical\n  - histopathology\n  - h-and-e\n  - nuclei\n  - nucleus-segmentation\n  - instance-segmentation\n  - breast-cancer\n  - tcga\n  - tcga-brca\npretty_name: NuCLS (corrected single-rater)\nsize_categories:\n  - 1K<n<10K\n---\n\n# NuCLS — Corrected Single-Rater Subset\n\n**Nucleus Classification, Localization and Segmentation** in breast cancer\n(Amgad et al., *GigaScience* 2022). H&E fields-of-view (FOVs) cropped from\n**TCGA-BRCA** whole-slide images at 40× (~0.2 µm/px), with per-nucleus instance +\nsemantic masks and a 13/7/4-class cell-type scheme.\n\n> **Scope — please read.** This repository contains **only the corrected\n> (pathologist-QC'd) single-rater** subset: **1,744 FOVs / 59,485 nuclei** from\n> **124 TCGA-BRCA slides**. It does **not** include the *uncorrected* single-rater\n> subset (2,168 FOVs) nor the *multi-rater* evaluation/P-truth sets (the paper's\n> gold unbiased-evaluation tier) — both remain at the official\n> [Google Drive](https://sites.google.com/view/nucls). The corrected subset is the\n> QC'd tier most downstream segmentation work uses, and the only one distributed as\n> a single stable archive.\n\n- **Modality:** Histopathology — H&E brightfield, variable-size FOVs at 40× (~0.2 µm/px)\n- **Target:** nucleus segmentation + classification (breast)\n- **Ground truth:** ready-made 3-channel instance+semantic masks (see encoding below)\n- **License:** **CC0 1.0 Universal** (the dataset is public domain; the paper text is CC BY 4.0)\n- **Source:** [official portal](https://sites.google.com/view/nucls) ·\n  [GitHub `PathologyDataScience/NuCLS`](https://github.com/PathologyDataScience/NuCLS)\n\n## ⚠️ Hybrid bounding-box + segmentation dataset\n\nOnly **33.1%** of nuclei (19,680 / 59,485) carry **true polygon boundaries**\n(`type == 'polyline'`). The other **66.7%** (39,694) are **bounding boxes**\n(`type == 'rectangle'`) **rasterized as filled rectangles** in the mask. The mask\nPNG does **not** distinguish them — the per-FOV `type` column (preserved in\n`annotations_csv`) is the only way to tell true contours from box approximations.\nAccount for this before treating every mask object as a pixel-accurate nucleus.\n\n## Mask encoding (`mask`, 3-channel uint8 PNG)\n\n| Channel | Meaning |\n|---|---|\n| **ch0 (R)** | semantic `GT_code` |\n| **ch1 (G)** | ROI/foreground flag (1 = inside annotated FOV ROI, 0 = outside; `G==1 ⇔ R>0`) |\n| **ch2 (B)** | instance id (0 = outside ROI, **2 = ROI-filler region**, 3..N = individual nuclei) |\n\nSemantic `GT_code` (ch0):\n\n| Code | Class | | Code | Class |\n|--|--|--|--|--|\n| 0 | background (outside ROI) | | 8 | myoepithelium |\n| 1 | tumor | | 9 | apoptotic_body |\n| 2 | fibroblast | | 10 | neutrophil |\n| 3 | lymphocyte | | 11 | ductal_epithelium |\n| 4 | plasma_cell | | 12 | eosinophil |\n| 5 | macrophage | | 99 | unlabeled-nucleus |\n| 6 | mitotic_figure | | 253 | ROI-region (non-nucleus tissue) |\n| 7 | vascular_endothelium | | | |\n\n```python\nimport numpy as np\nm = np.array(row[\"mask\"])          # (H, W, 3) uint8 — read via numpy, do NOT convert(\"L\")\nnucleus = (m[..., 0] >= 1) & (m[..., 0] <= 99)   # binary nucleus mask (excludes bg & ROI-region)\ninst    = m[..., 2]                # per-nucleus instance ids (ROI-filler id has R==253)\nsem     = m[..., 0]               # semantic class codes (table above)\n```\n\nThe 13-class `raw` codes group into the 7 `main` and 4 `super` classes used in the\npaper; the per-nucleus `raw`/`main`/`super` labels are in `annotations_csv`.\n\n## Columns\n\n| Column | Type | Notes |\n|---|---|---|\n| `fov_id` | string | FOV filename stem (encodes slide barcode + WSI crop coords) |\n| `image` | Image (RGB) | H&E FOV, cropped to align with `mask` |\n| `mask` | Image (RGB, 3-ch uint8) | instance+semantic mask, encoding above; cropped to match `image` |\n| `slide_name` | string | **TCGA slide barcode** (e.g. `TCGA-A1-A0SP-DX1`) — cross-dataset dedup key |\n| `tcga_patient` | string | TCGA patient barcode (e.g. `TCGA-A1-A0SP`) |\n| `hospital` | string | TCGA TSS code (e.g. `A1`) — the slide-grouping used b","encoding":"utf8","key":"README.md","name":"README.md","size":6658,"truncated":true}],"sourceObjectKey":"data/train-00000-of-00003.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/NuCLS\")\nprint(ds)","starterSamplePath":"data/train-00000-of-00003.parquet","totalObjects":5},"summary":"NuCLS (Nucleus Classification, Localization and Segmentation) is a crowdsourced and pathologist-reviewed annotation set of cell nuclei in breast cancer histology.","title":"NuCLS: Nucleus Classification, Localization and Segmentation in TCGA breast cancer H&E (corrected single-rater set, 59k nuclei)","updatedAt":"2026-11-22T00:57:57.067Z","visibility":"public","websiteUrl":"https://sites.google.com/view/nucls"}}