README.md

---
license: cc-by-nc-sa-4.0
pretty_name: Pathomics

# task_categories:
#   - feature-extraction
#   - image-classification

# tags:
#   - pathology
#   - computational-pathology
#   - spatial-transcriptomics
#   - whole-slide-images
#   - multimodal
#   - foundation-model

# language:
#   - en

# size_categories:
#   - 1K<n<10K
# ---

# # Pathomics

# Pathomics is a large-scale multimodal pathology dataset integrating:

# - Whole-slide pathology images (WSIs)
# - Spatial transcriptomics (ST)
# - Tissue metadata
# - Visualization assets
# - Foundation-model-ready preprocessing outputs

# The dataset extends the original HEST dataset with:

# - CellViT-based nuclei segmentation
# - Additional curated spatial transcriptomics datasets from literature
# - Standardized multimodal organization
# - Unified metadata schema for downstream AI applications

# Pathomics is designed for:

# - Computational pathology
# - Spatial transcriptomics research
# - Vision-language foundation models
# - Cell-level representation learning
# - Multimodal biomedical AI

# ---

# # Relationship with HEST

# Pathomics contains two types of samples:

# | source | Description |
# |---|---|
# | `hest` | Samples originating from the original HEST dataset |
# | `literature` | Additional curated samples processed independently |

# For samples with:

# ```text
# source = hest
# ````

# the corresponding base HEST data can optionally be downloaded automatically.

# Pathomics stores:

# * Cell segmentation results
# * Additional metadata
# * Standardized file organization
# * Derived multimodal assets

# while HEST provides:

# * Original WSI/ST assets
# * Base preprocessing outputs

# ---

# # Dataset Structure

# The repository is organized by modality/type instead of per-sample folders.

# ```text
# pathomics/

# ├── metadata/
# │   └── NCBI689.json
# │
# ├── st/
# │   └── NCBI689.h5ad
# │
# ├── wsis/
# │   └── NCBI689.tif
# │
# ├── thumbnails/
# │   └── NCBI689_downscaled_fullres.jpeg
# │
# ├── spatial_plots/
# │   └── NCBI689_spatial_plots.png
# │
# ├── cellvit_seg_for_superfocus/
# │   └── NCBI689/
# │       └── ...
# │
# ├── PATHOMICS_v3_0_0.csv
# │
# └── README.md
# ```

# ---

# # File Descriptions

# | Directory                     | Description                                    |
# | ----------------------------- | ---------------------------------------------- |
# | `metadata/`                   | JSON metadata for each sample                  |
# | `st/`                         | Spatial transcriptomics AnnData (`.h5ad`)      |
# | `wsis/`                       | Whole-slide pathology images                   |
# | `thumbnails/`                 | Downscaled JPEG tissue thumbnails              |
# | `spatial_plots/`              | Visualization of spatial transcriptomics spots |
# | `cellvit_seg_for_superfocus/` | CellViT segmentation outputs                   |
# | `PATHOMICS_v3_0_0.csv`        | Master metadata table                          |

# ---

# # Metadata Table

# The file:

# ```text
# PATHOMICS_v3_0_0.csv
# ```

# contains the master metadata table for all samples.

# Important fields include:

# | Field                     | Description                             |
# | ------------------------- | --------------------------------------- |
# | `id`                      | Unique sample identifier                |
# | `source`                  | `hest` or `literature`                  |
# | `hest_id`                 | Original HEST sample ID (if applicable) |
# | `organ`                   | Tissue/organ source                     |
# | `species`                 | Species information                     |
# | `platform`                | Spatial transcriptomics platform        |
# | `nb_genes`                | Number of genes                         |
# | `spots_under_tissue`      | Number of tissue-covered spots          |
# | `pixel_size_um_estimated` | Estimated pixel resolution              |
# | `cellvit_seg`             | Number of segmented cells               |
# | `has_superfocus_seg`      | Whether CellViT segmentation exists     |

# ---

# # Access Requirements

# To use this dataset, you need access to:

# 1. Pathomics
# 2. HEST (optional but recommended for HEST-derived samples)

# ---

# # Step 1 — Request Access

# ## Pathomics

# Click:

# ```text
# Request Access
# ```

# at the top of this page.

# ---

# ## HEST

# Request access at:

# [https://huggingface.co/datasets/MahmoodLab/hest](https://huggingface.co/datasets/MahmoodLab/hest)

# Access is automatically granted.

# ---

# # Step 2 — Create a Hugging Face Token

# Create a Hugging Face token at:

# [https://huggingface.co/settings/tokens](https://huggingface.co/settings/tokens)

# Recommended permission:

# * `Write`

---

# Step 3 — Install Dependencies

```bash
pip install huggingface-hub pandas
```

---

# Step 4 — Login

```python
from huggingface_hub import login

login(token="YOUR_HF_TOKEN")
```

---

<!-- # Download API

The following helper script provides a unified interface for downloading:

* Individual samples
* Multiple samples
* Entire dataset
* Specific modalities only
* Optional HEST base data
 -->
---

# Download Script

```python
from huggingface_hub import snapshot_download
import pandas as pd
import os


def download_pathomics(
    ids=None,
    pathomics_dir="pathomics_data",
    hest_dir="pathomics_data",
):

    # -----------------------------
    # load metadata index
    # -----------------------------
    meta = None
    
    csv_files = [f for f in os.listdir(pathomics_dir)
                 if f.startswith("PATHOMICS_v")]

    # print(csv_files)

    if len(csv_files) > 0:
        csv_path = os.path.join(pathomics_dir, sorted(csv_files)[-1])
        meta = pd.read_csv(csv_path)

    if ids is None:
        snapshot_download(
            repo_id="Boyoungc/pathomics",
            allow_patterns="*",
            repo_type="dataset",
            local_dir=pathomics_dir,
        )
        return

    # -----------------------------
    # split ids
    # -----------------------------
    hest_ids = []
    local_ids = []

    if meta is not None and "source" in meta.columns:
        sub = meta[meta["id"].isin(ids)]
        hest_ids = sub[sub["source"] == "hest"]["id"].tolist()
        local_ids = sub[sub["source"] != "hest"]["id"].tolist()
    else:
        local_ids = ids

    # =========================================================
    # 1. HEST DOWNLOAD (STRICT MODALITY FILTER)
    # =========================================================
    if len(hest_ids) > 0:

        hest_patterns = []
    
        for hid in hest_ids:
            hest_patterns.extend([
                f"metadata/{hid}.json",
                f"st/{hid}.h5ad",
                f"wsis/{hid}.tif",
                f"thumbnails/{hid}_*",
                f"spatial_plots/{hid}_*",
            ])
    
        snapshot_download(
            repo_id="MahmoodLab/hest",
            allow_patterns=hest_patterns,
            repo_type="dataset",
            local_dir=hest_dir,
        )
    
        print(f"[HEST] downloaded {len(hest_ids)} samples")

    # =========================================================
    # 2. PATHOMICS SEG ONLY for HEST
    # =========================================================
    if len(hest_ids) > 0:

        seg_patterns = [
            f"cellvit_seg_for_superfocus/{hid}/**"
            for hid in hest_ids
        ]

        snapshot_download(
            repo_id="Boyoungc/pathomics",
            allow_patterns=seg_patterns,
            repo_type="dataset",
            local_dir=pathomics_dir,
        )

        print(f"[SEG] downloaded HEST segmentations")

    # =========================================================
    # 3. PATHOMICS FULL for literature
    # =========================================================
    if len(local_ids) > 0:

        patterns = []

        for sid in local_ids:
            patterns.extend([
                f"metadata/{sid}.json",
                f"st/{sid}.h5ad",
                f"wsis/{sid}.tif",
                f"thumbnails/{sid}_*",
                f"spatial_plots/{sid}_*",
                f"cellvit_seg_for_superfocus/{sid}/**",
            ])

        snapshot_download(
            repo_id="Boyoungc/pathomics",
            allow_patterns=patterns,
            repo_type="dataset",
            local_dir=pathomics_dir,
        )

        print(f"[PATHOMICS] downloaded literature samples")
```

---

# Usage Examples

## Download One Sample

```python
download_pathomics(
    ids=["NCBI689"]
)
```

---

## Download Multiple Samples

```python
download_pathomics(
    ids=["NCBI689", "MEND62"]
)
```

---

## Download Entire Dataset

```python
download_pathomics()
```

---

## Download Only ST Data

```python
download_pathomics(
    ids=["NCBI689"],
    modalities=["st"]
)
```

---

<!-- # Acknowledgements

* HEST
* CellViT
* Hugging Face
* Spatial transcriptomics community
* Original data contributors -->

```
```