README.md

---
license: mit
dataset_info:
  features:
  - name: image
    dtype: image
  - name: slide_name
    dtype: string
  - name: x
    dtype: int64
  - name: y
    dtype: int64
  - name: level
    dtype: int64
  - name: patch_size
    sequence: int64
  - name: resize
    sequence: int64
  - name: embedding_vector
    sequence:
      sequence: float64
  splits:
  - name: train
    num_bytes: 4289155572.625
    num_examples: 40115
  download_size: 4289233317
  dataset_size: 4289155572.625
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# Dataset Card for Histopathology Dataset

## Dataset Summary
This dataset contains 1024x1024 patches of a group of histopathology images taken from the [CAMELYON16](http://gigadb.org/dataset/100439) dataset and embedding vectors extracted from these patches using the [Google Path Foundation](https://huggingface.co/google/path-foundation) model.

![Data Processing](data_processing.png)

## Thumbnail of Main Slide

![Main Slide Thumbnail](test_001.png)

## Usage

- [CAMELYON16](http://gigadb.org/dataset/100439): List of images taken from CAMELYON16 dataset:
    * `test_001.tif`
    * `test_002.tif`
    * `test_003.tif`
    * `test_004.tif`
    * `test_005.tif`
    * `test_006.tif`
    * `test_007.tif`
    * `test_008.tif`
    * `test_009.tif`
    * `test_010.tif`
    * `test_011.tif`
    * `test_012.tif`
    * `test_013.tif`
    * `test_014.tif`
    * `test_015.tif`
    * `test_016.tif`
    * `test_017.tif`
    * `test_018.tif`
    * `test_019.tif`
    * `test_020.tif`

* Create a folder named ***wsi*** and download the images given in the list. Then:

```python
from datasets import load_dataset
import openslide as ops


dataset = load_dataset("Cilem/histopathology-1024")
slide_name = dataset['train'][0]['slide_name']
resize = dataset['train'][0]['resize']
path = os.path.join('wsi', slide_name)
slide = ops.OpenSlide(path)
patch = slide.read_region((x, y), 0, patch_size)
patch = patch.resize(resize)
display(patch)
```


## Supported Tasks
Machine learning applications that can be performed using this dataset:
* Classification
* Segmentation
* Image generation

## Languages
* English

## Dataset Structure
### Data Fields
- `image`: Patch image.
- `slide_name`: Main slide name of the patch.
- `x`: X coordinate of the patch.
- `y`: Y coordinate of the patch.
- `level`: Level of the main slide.
- `patch_size`: Size of the patch.
- `resize`: Image size used to obtain embedding vector with Path foundation model.
- `embedding_vector`: Embedding vector of the patch extracted using Path foundation model.

## Dataset Creation
### Source Data
- **Original Sources**
  - [CAMELYON16](http://gigadb.org/dataset/100439): List of images taken from CAMELYON16 dataset:
    * `test_001.tif`
    * `test_002.tif`
    * `test_003.tif`
    * `test_004.tif`
    * `test_005.tif`
    * `test_006.tif`
    * `test_007.tif`
    * `test_008.tif`
    * `test_009.tif`
    * `test_010.tif`
    * `test_011.tif`
    * `test_012.tif`
    * `test_013.tif`
    * `test_014.tif`
    * `test_015.tif`
    * `test_016.tif`
    * `test_017.tif`
    * `test_018.tif`
    * `test_019.tif`
    * `test_020.tif`
  - [Google Path Foundation](https://huggingface.co/google/path-foundation): Embedding vectors extracted from the patches using the Path Foundation model.

## Considerations for Using the Data
### Social Impact and Bias
Attention should be paid to the Path Foundation model licenses provided by Google.