README.md

---
license: cc-by-nc-sa-4.0
task_categories:
  - image-classification
  - feature-extraction
language:
  - en
tags:
  - spatial-transcriptomics
  - multimodal
  - histology
  - gene-expression
  - lung
  - xenium
  - biology
  - single-cell
pretty_name: "CAMEO-Lung: A Multimodal Benchmark Dataset of Aligned H&E Patches and Gene Expression Profiles in the Lung"
size_categories:
  - 10K<n<100K
viewer: true
---

# CAMEO-Lung: A Multimodal Benchmark Dataset of Aligned H&E Patches and Gene Expression Profiles in the Lung

## Citation

If you use this dataset, please cite it directly and the original lung study:

```bibtex
@dataset{kuijs_cameo_lung_2026,
  author    = {Kuijs, Merel and Richter, Till and Gindra, Rushin H and Traeuble, Korbinian and
               Matek, Christian and Lukn{\'a}rov{\'a}, Rebeka and Peng, Tingying and Theis, Fabian J},
  title     = {CAMEO-Lung: A Multimodal Benchmark Dataset of Aligned H&E Patches and Gene Expression Profiles in the Lung},
  year      = {2026},
  publisher = {Hugging Face},
  doi       = {10.57967/hf/7910},
  url       = {https://huggingface.co/datasets/theislab/CAMEO-Lung}
}

@article{lpf,
  title   = {Spatial transcriptomics identifies molecular niche dysregulation associated with distal lung remodeling in pulmonary fibrosis},
  author  = {Vannan, Annika and Lyu, Ruqian and Williams, Arianna L and Negretti, Nicholas M and Mee, Evan D and Hirsh, Joseph and Hirsh, Samuel and Hadad, Niran and Nichols, David S and Calvi, Carla L and others},
  journal = {Nature Genetics},
  volume  = {57},
  number  = {3},
  pages   = {647--658},
  year    = {2025}
}
```

---

## Dataset Description

This dataset is part of the **CAMEO** framework for multimodal spatial transcriptomics learning. It contains paired histology images and gene expression data derived from the **Lung Pulmonary Fibrosis (LungPF)** 10x Xenium cohort, comprising **23 samples from 19 patients** covering healthy lung tissue and varying severities of pulmonary fibrosis.

Each row represents one **niche** — a 224×224 pixel crop of an H&E-stained histology slide paired with the single-cell gene expression profiles of all cells located within that crop, together with expert pathologist niche annotations, per-cell coordinates, and cell-type composition. In total, the dataset contains **71,309 niches** encompassing approximately **1 million cells** across 23 samples. We constructed these niche-level paired representations by spatially aligning the histological and transcriptomic modalities using [SpatialData](https://github.com/scverse/spatialdata), tessellating non-overlapping crops across each slide, and applying a quality control filter to exclude niches with less than 50% tissue coverage. Transcripts from partially included cells are treated as whole-cell data within the niche. Broad cell-type annotations (10 categories) and niche-level annotations were both provided by the original authors.

In addition to raw modality data, the dataset includes a set of **precomputed embeddings** from several unimodal foundation models to facilitate research on multimodal and unimodal representation learning.

- **Organization:** [Theislab](https://huggingface.co/theislab)
- **Source data:** Lung Pulmonary Fibrosis (LungPF) Xenium cohort
- **License:** [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/)

---

## Dataset Structure

### Splits

The dataset is stored as a single `train` split containing all 71,309 niches across all 23 samples.

| Split | Niches |
|-------|--------|
| Full dataset (`train`) | 71,309 |

---

## Column Descriptions

Each row corresponds to one niche (224×224 px patch). The following columns are included:

### Identifiers and labels

| Column | Type | Description |
|--------|------|-------------|
| `name` | `string` | Sample (slide) identifier, e.g. `"VUILD102LF"`. Maps to one of the 23 Xenium samples. |
| `annotation` | `ClassLabel` (int64) | Expert pathologist niche annotation, encoded as an integer. See [Niche Label Mapping](#niche-label-mapping) below. |
| `species` | `ClassLabel` (int64) | Species label. Always `0` (human) in this cohort. |
| `cancer` | `ClassLabel` (int64) | Cancer flag. Always `0` (False) in this cohort. |
| `tissue` | `ClassLabel` (int64) | Tissue label. Always `0` (lung) in this cohort. |

### Raw modality data

| Column | Type | Shape | Description |
|--------|------|-------|-------------|
| `image` | `Image` | 224×224 RGB | H&E-stained histology patch. |
| `gexp` | `Array2D` float32 | (200, 343) | Raw gene expression counts per cell. Up to 200 cells per niche (zero-padded); 343 Xenium panel genes. Use `mask` to identify valid cells. |
| `spot_gexp` | `Array2D` float32 | (1, 343) | Niche-level pseudobulk gene expression (sum over valid cells). |
| `mask` | `Sequence` bool | (200,) | Boolean mask indicating valid cells (`True` = real cell, `False` = padding). |
| `cell_coords` | `Array2D` int32 | (200, 2) | Cell centroid coordinates (x, y) in pixel space within the 224×224 patch. Padded to 200 rows. |
| `cell_type_ratio` | `Sequence` float32 | (10,) | Fraction of each of the 10 broad cell types present in the niche. |

### Precomputed embeddings

All embeddings are niche-level representations derived from the raw modalities.

| Column | Type | Shape | Description |
|--------|------|-------|-------------|
| `img_embed` | `Sequence` float32 | (1024,) | Image embedding from [UNI](https://github.com/mahmoodlab/UNI) |
| `conch_embedding` | `Sequence` float64 | (512,) | Image embedding from [CONCH](https://github.com/mahmoodlab/CONCH) |
| `ctranspath_embedding` | `Sequence` float64 | (768,) | Image embedding from [CTransPath](https://github.com/Xiyue-Wang/TransPath). |
| `gexp_embed` | `Sequence` float32 | (128,) | Gene expression embedding learned by a self-supervised Graph Attention Network. |
| `scvi_pool` | `Sequence` float64 | (128,) | [scVI](https://scvi-tools.org/) embedding, pooled over valid cells in the niche. |
| `scvi_pseudobulk` | `Sequence` float64 | (128,) | scVI embedding computed from the pseudobulk niche expression profile. |
| `pca_pool` | `Sequence` float64 | (128,) | PCA embedding (128 components), pooled over valid cells in the niche. |
| `pca_pseudobulk` | `Sequence` float64 | (128,) | PCA embedding computed from the pseudobulk niche expression profile. |
| `nicheformer_pool` | `Sequence` float64 | (512,) | [Nicheformer](https://github.com/theislab/nicheformer) embedding, pooled over valid cells in the niche. |
| `scgpt_pool` | `Sequence` float64 | (512,) | [scGPT](https://github.com/bowang-lab/scGPT) embedding, pooled over valid cells in the niche. |

### Niche Label Mapping

The `annotation` column contains integer class labels corresponding to expert-annotated niche types:

| Integer | Niche type |
|---------|------------|
| 0 | Advanced Remodeling |
| 1 | Airway Smooth Muscle |
| 2 | Artery |
| 3 | Emphysema |
| 4 | Fibroblastic Focus |
| 5 | Fibrosis |
| 6 | Giant Cell |
| 7 | Goblet Cell Metaplasia |
| 8 | Granuloma |
| 9 | Hyperplastic AECs |
| 10 | Interlobular Septum |
| 11 | Large Airway |
| 12 | Microscopic Honeycombing |
| 13 | Minimally Remodeled Alveoli |
| 14 | Mixed Inflammation |
| 15 | Muscularized Artery |
| 16 | NOANNOT |
| 17 | Normal Alveoli |
| 18 | Remnant Alveoli |
| 19 | Remodeled Epithelium |
| 20 | Severe Fibrosis |
| 21 | Small Airway |
| 22 | TLS |
| 23 | Venule |

---

## Loading the Dataset

### Standard loading

```python
from datasets import load_dataset

dataset = load_dataset("theislab/CAMEO-Lung")
train_ds = dataset["train"]

# Access one example
example = train_ds[0]
print(example.keys())
# dict_keys(['name', 'image', 'img_embed', 'gexp_embed', 'cell_type_ratio',
#            'annotation', 'species', 'cancer', 'tissue', 'gexp', 'mask',
#            'cell_coords', 'spot_gexp', 'conch_embedding', 'ctranspath_embedding',
#            'pca_pool', 'pca_pseudobulk', 'scvi_pool', 'scvi_pseudobulk',
#            'nicheformer_pool', 'scgpt_pool'])

# The image is a PIL Image
print(example["image"].size)       # (224, 224)

# Gene expression: shape (200, 343), use mask to select valid cells
import numpy as np
gexp = np.array(example["gexp"])         # shape (200, 343)
mask = np.array(example["mask"])         # shape (200,) bool
gexp_valid = gexp[mask]                  # shape (n_cells, 343)

# Decode the niche label
label_name = train_ds.features["annotation"].int2str(example["annotation"])
print(label_name)  # e.g. "Normal Alveoli"
```

### Streaming (avoids downloading all ~25 GB upfront)

```python
from datasets import load_dataset

dataset = load_dataset("theislab/CAMEO-Lung", streaming=True)
for example in dataset["train"]:
    # process one niche at a time
    break
```

### Filtering by sample

```python
train_samples = ["VUILD102LF", ...]
train_split = dataset["train"].filter(lambda x: x["name"] in train_samples)
```

---

## License

This dataset is distributed under the [Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.