---
license: cc-by-4.0
task_categories:
- video-classification
tags:
- medical
- ultrasound
- eye
- ocular
- classification
arxiv: 2508.04735
size_categories:
- 1K<n<10K
language:
- en
pretty_name: erdes
dataset_info:
  features:
  - name: clip_id
    dtype: string
  - name: file_path
    dtype: string
  - name: diagnostic_class
    dtype: string
  - name: subtype
    dtype: string
  - name: anatomical_subclass
    dtype: string
  - name: fps
    dtype: float32
  - name: frame_count
    dtype: int32
  - name: width
    dtype: int32
  - name: height
    dtype: int32
  - name: duration_seconds
    dtype: float32
  splits:
  - name: train
    num_examples: 5381
configs:
- config_name: default
  data_files:
  - split: train
    path: metadata.csv
- config_name: non_rd_vs_rd
  data_files:
  - split: train
    path: splits/non_rd_vs_rd/train.csv
  - split: validation
    path: splits/non_rd_vs_rd/val.csv
  - split: test
    path: splits/non_rd_vs_rd/test.csv
- config_name: macula_detached_vs_intact
  data_files:
  - split: train
    path: splits/macula_detached_vs_intact/train.csv
  - split: validation
    path: splits/macula_detached_vs_intact/val.csv
  - split: test
    path: splits/macula_detached_vs_intact/test.csv
- config_name: normal_vs_pvd
  data_files:
  - split: train
    path: splits/normal_vs_pvd/train.csv
  - split: validation
    path: splits/normal_vs_pvd/val.csv
  - split: test
    path: splits/normal_vs_pvd/test.csv
- config_name: normal_vs_rd
  data_files:
  - split: train
    path: splits/normal_vs_rd/train.csv
  - split: validation
    path: splits/normal_vs_rd/val.csv
  - split: test
    path: splits/normal_vs_rd/test.csv
- config_name: pvd_vs_rd
  data_files:
  - split: train
    path: splits/pvd_vs_rd/train.csv
  - split: validation
    path: splits/pvd_vs_rd/val.csv
  - split: test
    path: splits/pvd_vs_rd/test.csv
---
# ERDES: Eye Retinal Detachment Ultrasound Dataset

| Resource | Link |
|----------|------|
| Website | [![Website](https://img.shields.io/badge/Website-ERDES-green)](https://osupcvlab.github.io/ERDES/) |
| Paper | [![arXiv](https://img.shields.io/badge/arXiv-2508.04735-b31b1b.svg)](https://arxiv.org/abs/2508.04735) |
| Checkpoints | [![HF Checkpoints](https://img.shields.io/badge/🤗-Checkpoints-yellow)](https://huggingface.co/collections/pcvlab/erdes-ocular-ultrasound-classification-all-tasks) [![Zenodo](https://img.shields.io/badge/Zenodo-Checkpoints-blue)](https://zenodo.org/records/18821031) |
| Dataset | [![Zenodo](https://img.shields.io/badge/Zenodo-Dataset-blue)](https://zenodo.org/records/18644370) |
| Code | [![GitHub](https://img.shields.io/badge/GitHub-OSUPCVLab/ERDES-black?logo=github)](https://github.com/OSUPCVLab/ERDES) |

## 📌 Introduction
**ERDES** is a large-scale, publicly available dataset of **3D ocular ultrasound videos** for retinal and macular detachment classification. It was introduced in our paper **[ERDES: A Benchmark Video Dataset for Retinal Detachment and Macular Status Classification in Ocular Ultrasound](https://arxiv.org/abs/2508.04735)**. The corpus consists of **5,381 expertly annotated video clips** totaling **5 hours and 10 minutes**, providing a valuable resource for medical AI research in ophthalmology.

### Key Features:
- **5,381 labeled ultrasound video clips**
- **Expert annotations for retinal detachment (RD) and macular status**
- **Structured classification (Normal, RD, PVD, macula-detached/intact)**
- **Preprocessed for privacy and consistency**  

---

## 🎯 Motivation  
Medical video datasets for AI are scarce despite their clinical importance. ERDES bridges this gap by offering:  
- A **standardized benchmark** for retinal detachment classification in ultrasound videos.  
- Support for **spatiotemporal analysis** (e.g., 3D CNNs).  
- Open access to accelerate research in **ocular diagnostics**.  

---

## 📊 Dataset Overview  

### **1. Data Structure**
Videos are categorized into two primary groups:

**Non-RD (Non-Retinal Detachment):**
- Normal
- Posterior Vitreous Detachment (PVD)

**RD (Retinal Detachment):**
- Macula-Detached
  - Bilateral (nasal and temporal regions involved)
  - Temporal detachment only
- Macula-Intact
  - Nasal detachment
  - Temporal detachment

<div align="center">
<img src="media/folder_structure.png" alt="Folder Structure" width="150" />
</div>

### **2. Annotations**  
Each clip is labeled by sonologists for:  
- Presence/absence of retinal detachment.  
- Macular involvement (detached/intact).  

<div align="center">
<img src="media/labeling_rd.png" alt="Data Statistics" width="500" />
</div>

### **3. Preprocessing**
- **Privacy**: PHI removed using YOLOv8-based globe detection.
- **Consistency**: Cropped to the ocular ROI.
- **Format**: MP4 videos.

---

## 📥 Download  
Access the dataset via the HuggingFace API:

```py
from datasets import load_dataset

dataset = load_dataset("pcvlab/erdes")
```

---

## 🛠️ Code & Baselines  
We open source our baseline experiments on our **[GitHub](https://github.com/OSUPCVLab/ERDES)** repo, which includes:  

- Baseline 3D CNN and ViT models for classification.
- End-to-end diagnostic pipeline for macular detachment.   

---

## 📜 Citation
If you use ERDES, please cite:
```bibtex
@article{ozkuterdes,
  title={ERDES: A Benchmark Video Dataset for Retinal Detachment and Macular Status Classification in Ocular Ultrasound},
  author={Ozkut, Yasemin and Navard, Pouyan and Adhikari, Srikar and Situ-LaCasse, Elaine and Acu{\~n}a, Josie and Yarnish, Adrienne A and Yilmaz, Alper},
  journal={arXiv preprint arXiv:2508.04735},
  year={2025}
}
```