{"success":true,"data":{"author":"Parmida Ghahremani, Saad Nadeem et al. (Memorial Sloan Kettering); mirrored by MedOtter","bucketAccessType":"public","bucketName":"MedOtter/DeepLIIF","bucketPrefix":"","category":"Health & Biomedical","cloudProvider":"huggingface","createdAt":"2026-11-21T20:45:56.045Z","description":"DeepLIIF (Deep Learning-Inferred Multiplex ImmunoFluorescence) pairs brightfield Ki67-DAB immunohistochemistry (IHC) patches with co-registered multiplex immunofluorescence (mpIF) of the same tissue, so a model can learn to predict fluorescence channels and score Ki67-positive versus Ki67-negative cells from an ordinary IHC stain. Each 512x512 sample has six aligned panels: the IHC input, a reconstructed hematoxylin channel, mpIF DAPI (nuclear stain), mpIF Lap2 (nuclear envelope), the mpIF Ki67 marker, and a ground-truth mask in which red marks Ki67-positive cells, blue Ki67-negative cells and green cell boundaries. There are 1,715 samples: 575 train, 91 validation and 598 test from bladder and lung (DeepLIIF), plus 385 train and 66 validation breast carcinoma patches (BC-DeepLIIF). Parquet files under data/ with images as PNG bytes, about 2.5 GB (measured 2026-11-21); read with Hugging Face datasets or pandas plus Pillow. Use for virtual immunofluorescence staining, cell segmentation and Ki67 proliferation index quantification in digital pathology. Mirrored by MedOtter from the authors' Zenodo record 4751737, which is released under CC BY 4.0; cite Ghahremani et al., Nature Machine Intelligence 2022 (doi 10.1038/s42256-022-00471-x).","displayTitle":"DeepLIIF: Ki67 immunohistochemistry (IHC) with co-registered multiplex immunofluorescence (mpIF) and cell masks, bladder, lung and breast","huggingFaceUrl":"https://huggingface.co/datasets/MedOtter/DeepLIIF","id":"b265e428-0d93-59bf-b347-0db9e0c483c0","license":"CC-BY-4.0","licenseEvidenceUrl":"https://zenodo.org/records/4751737","objectCountEstimate":17,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1038/s42256-022-00471-x","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/DeepLIIF\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"MedOtter/DeepLIIF\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=MedOtter, license:cc-by-4.0, size_categories:1K<n<10K, format:parquet, downloads=1615"],"format":"mixed","generatedAt":"2026-11-21T20:48:38.708Z","layout":{"example":"data/bc_train-00000-of-00003.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"50415231150415e80715a4044c1580011500120000f40318020000003131040506045f300d08111000320d1000330d0800340d0800350d0800360d0800370d0800380d0800390556091600370d5600370d5600370d5600370d5600370d5600370d5600370d5600370d5600370d5600370d56096400380d5600380d5600380d5600380d5600380d5600380d5600380d5600380d5600380d5600380d56003905ba00390d5600390d5600390d5600390d5600390d5600390d5600390d5600390d5600390d56003909560032254800322d58043231095604323109560432310956043231095604323109560432310956043231095604323109560432310d56258e04323309ac0432330d5600330d5600330d5600330d5604335f25ae2c32335f360400000032335f371500157215762c158001151015061506000039e003000000800101061140200c44611c48a22c4ce33c50244d54655d58a66d5ce77d60288e64699e68aaae6cebbe702ccf746ddf78aeef7cefff1504151015144c15021500120000081c000000000000000015001516151a2c15800115101506150600000b2803000000800101018001001504151e15224c150215001200000f380b00000042432d446565704c49494615001516151a2c15800115101506150600000b280300000080010101800100150015b6b55e15b6b55e2c150415001506150600000200","encoding":"hex","key":"data/bc_train-00000-of-00003.parquet","truncatedAt":512}],"sampleObjects":["data/bc_train-00000-of-00003.parquet","data/bc_train-00001-of-00003.parquet","data/bc_train-00002-of-00003.parquet","data/bc_validation-00000-of-00001.parquet","data/test-00000-of-00005.parquet"],"sidecars":[{"bytes":"---\nlicense: cc-by-4.0\ntask_categories:\n  - image-segmentation\ntags:\n  - medical\n  - histopathology\n  - immunohistochemistry\n  - ihc\n  - multiplex-immunofluorescence\n  - ki67\n  - cell-segmentation\n  - deepliif\npretty_name: DeepLIIF\nsize_categories:\n  - 1K<n<10K\n---\n\n# DeepLIIF (Deep Learning-Inferred Immunofluorescence)\n\nCo-registered IHC (Ki67-DAB brightfield) and multiplex immunofluorescence (mpIF)\npatches with cell-level segmentation + classification ground truth for\nquantification of clinical pathology slides.\n\nSource paper: Ghahremani et al., \"Deep learning-inferred multiplex\nimmunofluorescence for immunohistochemical image quantification,\"\n*Nature Machine Intelligence* 4(4):401-412, 2022.\nDOI: [10.1038/s42256-022-00471-x](https://doi.org/10.1038/s42256-022-00471-x).\nZenodo: <https://zenodo.org/records/4751737>.\n\n## Overview\n- **Modality:** Histopathology - IHC brightfield + co-registered mpIF (DAPI, Lap2, Ki67 marker)\n- **Patch size:** 512x512 each (originals are 3072x512 PNGs concatenating six panels)\n- **Tissue:** Bladder + Lung (main DeepLIIF), Breast carcinoma (BC-DeepLIIF subset)\n- **Total samples:** 1,715\n  - `train` (DeepLIIF): 575\n  - `validation` (DeepLIIF): 91\n  - `test` (DeepLIIF): 598\n  - `bc_train` (BC-DeepLIIF, breast): 385\n  - `bc_validation` (BC-DeepLIIF, breast): 66\n\n## Columns\n| Column | Type | Notes |\n|---|---|---|\n| `sample_id` | string | Original PNG stem |\n| `tissue` | ClassLabel(3) | `0=BC`, `1=Bladder`, `2=Lung` |\n| `subset` | string | `DeepLIIF` (main) or `BC-DeepLIIF` |\n| `ihc` | Image (RGB) | Input - 512x512 brightfield Ki67-DAB IHC |\n| `hematoxylin` | Image (RGB) | Aux target - reconstructed hematoxylin channel |\n| `dapi` | Image (RGB) | Aux target - mpIF DAPI nuclear stain |\n| `lap2` | Image (RGB) | Aux target - mpIF Lap2 nuclear-envelope stain |\n| `marker` | Image (RGB) | Aux target - mpIF Ki67 marker channel |\n| `seg_mask` | Image (RGB) | **Ground truth** - red=Ki67+ cell, blue=Ki67- cell, green=boundary, black=background |\n\n## Ground Truth\nThe `seg_mask` column is the canonical GT. It was generated by combining\nper-modality instance segmentations from mpIF DAPI + Lap2 + Hematoxylin + IHC,\nwith instance boundaries initialized by the ImPartial interactive framework on\nDAPI. Cells are then classified red/blue based on Ki67 marker positivity.\n\nFor binary semantic segmentation (nucleus vs background) treat (red OR blue)\npixels as foreground. For positive-vs-negative classification, decode red and\nblue channels separately.\n\n## Derivation\nEach source PNG was sliced column-wise into six 512x512 panels:\n- Columns [0:512]       -> `ihc`\n- Columns [512:1024]    -> `hematoxylin`\n- Columns [1024:1536]   -> `dapi`\n- Columns [1536:2048]   -> `lap2`\n- Columns [2048:2560]   -> `marker`\n- Columns [2560:3072]   -> `seg_mask`\n\nNo other preprocessing.\n\n## License\nCC BY 4.0 (dataset, per Zenodo record 4751737). The DeepLIIF code repo is\nApache 2.0 with Commons Clause; that license applies only to code/models, not\nto this imaging data.\n\n## Citation\n```bibtex\n@article{ghahremani2022deep,\n  title={Deep learning-inferred multiplex immunofluorescence for\n         immunohistochemical image quantification},\n  author={Ghahremani, Parmida and Li, Yanyun and Kaufman, Arie and\n          Vanguri, Rami and Greenwald, Noah and Angelo, Michael and\n          Hollmann, Travis J and Nadeem, Saad},\n  journal={Nature Machine Intelligence},\n  volume={4}, number={4}, pages={401--412}, year={2022},\n  doi={10.1038/s42256-022-00471-x}\n}\n```\n","encoding":"utf8","key":"README.md","name":"README.md","size":3496,"truncated":false}],"sourceObjectKey":"data/bc_train-00000-of-00003.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/DeepLIIF\")\nprint(ds)","starterSamplePath":"data/bc_train-00000-of-00003.parquet","totalObjects":17},"summary":"DeepLIIF (Deep Learning-Inferred Multiplex ImmunoFluorescence) pairs brightfield Ki67-DAB immunohistochemistry (IHC) patches with co-registered multiplex immunofluorescence (mpIF)…","title":"DeepLIIF: Ki67 immunohistochemistry (IHC) with co-registered multiplex immunofluorescence (mpIF) and cell masks, bladder, lung and breast","updatedAt":"2026-11-22T00:52:46.835Z","visibility":"public","websiteUrl":"https://zenodo.org/records/4751737"}}