README.md

---
language:
- en
license: other
tags:
- medical
- ophthalmology
- fundus-image
- image-classification
- image-segmentation
- pathologic-myopia
- multi-task
- optic-disc
- lesion-segmentation
task_categories:
- image-classification
- image-segmentation
- object-detection
task_ids:
- multi-class-image-classification
- semantic-segmentation
pretty_name: PALM (Pathologic Myopia) Fundus Image Dataset
size_categories:
- 1K<n<10K
annotations_creators:
- expert-generated
source_datasets:
- original
source_data_urls:
- https://ieee-dataport.org/documents/palm-pathologic-myopia-challenge
---

# 🩺 PALM — Pathologic Myopia Fundus Image Dataset

<table align="center">
    <tr>
        <td width="100%" align="center">
            <img src="rm_images/Merged_Fundus_Images_with_Captions.jpg" alt="Merged Dataset Samples" style="max-width: 100%; height: auto;">
            <br>
            <p><strong>Image:</strong> Dataset Samples.</p>
        </td>
    </tr>
</table>

---

## 📘 Overview

**PALM (Pathologic Myopia)** is a publicly available fundus image dataset developed for detecting **pathologic myopia (PM)** and analyzing associated retinal lesions and anatomical structures.

It was released for the **Pathologic Myopia Challenge (PALM)**, hosted by the **Chinese Academy of Sciences** and **Sun Yat-sen University**, and published on **IEEE DataPort**.
The dataset provides both **classification** and **segmentation** tasks, making it valuable for multi-task ophthalmic AI research.

🔗 **Official Sources:**
- [PALM Challenge (IEEE DataPort)](https://ieee-dataport.org/documents/palm-pathologic-myopia-challenge)
- [PALM Dataset Publication (PMC)](https://pmc.ncbi.nlm.nih.gov/articles/PMC10799845/)
- [ResearchGate Dataset Overview](https://www.researchgate.net/figure/Folder-organization-of-our-PALM-dataset_fig6_377564136)

---

## 📊 Dataset Summary

| Feature | Description |
|----------|-------------|
| **Images** | 1,200 color fundus photographs |
| **Labels** | Binary — *Pathologic Myopia (PM)* or *Non-PM* |
| **Annotations** | Optic disc boundary, fovea location, and lesion masks (atrophy, detachment) |
| **Resolution** | Varies (45° field-of-view fundus images) |
| **Format** | JPEG |
| **Tasks** | Classification, Segmentation |
| **Source Institutions** | Multiple ophthalmic centers in China |
| **License** | Free for research and educational use |
| **Released** | 2019 (PALM Challenge) |

---

## 📁 Folder Structure

```python
PALM/
├── images/
│ ├── train/
│ ├── validation/
│ └── test/
├── annotations/
│ ├── optic_disc_masks/
│ ├── lesion_masks/
│ └── fovea_locations.csv
├── labels.csv
└── README.md
```

---

## 🩸 Labels and Annotations

- **Classification Label:**
  - `1` = Pathologic Myopia (PM)
  - `0` = Non-Pathologic (Normal)

- **Segmentation Annotations:**
  - Optic disc boundaries
  - Fovea coordinates
  - Lesion masks for:
    - Patchy atrophy
    - Retinal detachment
    - Peripapillary atrophy

---

## 🧠 Research Applications

PALM is designed for:

- **Pathologic Myopia detection** from fundus images
- **Segmentation** of optic disc, fovea, and lesion regions
- **Multi-task learning** combining classification and segmentation
- **Explainable AI** studies on high-myopia pathology

---

## ⚙️ Limitations

- Limited image count (~1,200)
- Variability in camera type and illumination
- Binary labeling (PM vs Non-PM) does not cover all clinical myopia subtypes
- Lesion annotations may need preprocessing for some segmentation frameworks

---

## 📥 Access and Citation

### 🔗 Access
Dataset available via official IEEE DataPort page:
👉 [https://ieee-dataport.org/documents/palm-pathologic-myopia-challenge](https://ieee-dataport.org/documents/palm-pathologic-myopia-challenge)

### 📄 Citation

If you use PALM, please cite:
Fang H., Li F., Wu J., Fu H., Sun X., Orlando J. I., Bogunović H., Zhang X., Xu Y.
PALM: Open Fundus Photograph Dataset with Pathologic Myopia Recognition and Anatomical Structure Annotation.
IEEE DataPort, 2019.

---

## 🧾 License

The PALM dataset is made available **for research and educational use only**.
Redistribution or commercial use requires permission from the dataset authors.