Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,55 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- image-classification
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
pretty_name: PANDAMIL
|
| 8 |
+
---
|
| 9 |
+
# PANDA - Multiple Instance Learning (MIL)
|
| 10 |
+
|
| 11 |
+
*Important.* This dataset is part of the [**torchmil** library](https://franblueee.github.io/torchmil/).
|
| 12 |
+
|
| 13 |
+
This repository provides an adapted version of the [Prostate cANcer graDe Assessment (PANDA) dataset](https://panda.grand-challenge.org/data/) tailored for **Multiple Instance Learning (MIL)**. It is designed for use with the [`PANDAMILDataset`](https://franblueee.github.io/torchmil/api/datasets/pandamil_dataset/) class from the [**torchmil** library](https://franblueee.github.io/torchmil/). PANDA is a widely used benchmark in MIL research, making this adaptation particularly valuable for developing and evaluating MIL models.
|
| 14 |
+
|
| 15 |
+
### Dataset Description
|
| 16 |
+
|
| 17 |
+
We have preprocessed the whole-slide images (WSIs) by extracting relevant patches and computing features for each patch using various feature extractors.
|
| 18 |
+
|
| 19 |
+
- A **patch** is labeled as positive (`patch_label=1`) if more than 50% of its pixels are annotated as cancerous.
|
| 20 |
+
- A **WSI** is labeled as positive (`label=1`) if it contains at least one positive patch.
|
| 21 |
+
|
| 22 |
+
This means a slide is considered positive if there is any evidence of cancerous tissue.
|
| 23 |
+
|
| 24 |
+
### Directory Structure
|
| 25 |
+
|
| 26 |
+
After extracting the contents of the `.tar.gz` archives, the following directory structure is expected:
|
| 27 |
+
|
| 28 |
+
```
|
| 29 |
+
root
|
| 30 |
+
├── patches_{patch_size}
|
| 31 |
+
│ ├── features
|
| 32 |
+
│ │ ├── features_{features_name}
|
| 33 |
+
│ │ │ ├── wsi1.npy
|
| 34 |
+
│ │ │ ├── wsi2.npy
|
| 35 |
+
│ │ │ └── ...
|
| 36 |
+
│ ├── labels
|
| 37 |
+
│ │ ├── wsi1.npy
|
| 38 |
+
│ │ ├── wsi2.npy
|
| 39 |
+
│ │ └── ...
|
| 40 |
+
│ ├── patch_labels
|
| 41 |
+
│ │ ├── wsi1.npy
|
| 42 |
+
│ │ ├── wsi2.npy
|
| 43 |
+
│ │ └── ...
|
| 44 |
+
│ ├── coords
|
| 45 |
+
│ │ ├── wsi1.npy
|
| 46 |
+
│ │ ├── wsi2.npy
|
| 47 |
+
│ │ └── ...
|
| 48 |
+
└── splits.csv
|
| 49 |
+
```
|
| 50 |
+
|
| 51 |
+
Each `.npy` file corresponds to a single WSI. The `splits.csv` file defines train/test splits for standardized experimentation.
|
| 52 |
+
|
| 53 |
+
### About the original PANDA Dataset
|
| 54 |
+
|
| 55 |
+
The original [PANDA dataset](https://panda.grand-challenge.org/data/) contains WSIs of hematoxylin and eosin (H&E) stained prostate biopsy samples. The task is to classify the severity of prostate cancer within each slide, and to localize the cancerous tissue precisely. The dataset includes high-quality pixel-level annotations marking the cancerous tissue.
|