# DeepLIIF: Ki67 immunohistochemistry (IHC) with co-registered multiplex immunofluorescence (mpIF) and cell masks, bladder, lung and breast

## Facts

- Provider: Parmida Ghahremani, Saad Nadeem et al. (Memorial Sloan Kettering); mirrored by MedOtter
- License: CC-BY-4.0 (https://zenodo.org/records/4751737)
- Size: 17 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/MedOtter/DeepLIIF
- Access: public, readable without an account
- Category: Health & Biomedical
- Source: https://doi.org/10.1038/s42256-022-00471-x
- Page: https://scigantic.com/archive/b265e428-0d93-59bf-b347-0db9e0c483c0
- Keywords: Health & Biomedical, image segmentation, Parquet, Hugging Face

## About

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).

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("MedOtter/DeepLIIF")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `data/bc_train-00000-of-00003.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/b265e428-0d93-59bf-b347-0db9e0c483c0
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/b265e428-0d93-59bf-b347-0db9e0c483c0
- Over MCP: `get_schema_card` with id `b265e428-0d93-59bf-b347-0db9e0c483c0` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Image segmentation datasets](https://scigantic.com/topics/image-segmentation)
- [Health & Biomedical datasets](https://scigantic.com/topics/health-and-biomedical)
