2D Anatomy Segmentation Datasets Repository

Welcome to the 2D Anatomy Segmentation Datasets repository! This project provides a collection of scripts, tools, and datasets to streamline training and evaluation for anatomical segmentation tasks in medical imaging. Whether you're a researcher or practitioner in the medical imaging field, this repository is designed to collect 2D anatomical datasets and make it easy to download, process, and visualize data from various imaging modalities.

---

📦 Datasets Included

This repository supports the following publicly available datasets:

Dataset	Modality	#Images	Description	Link
BS80k	Scintigraphy	6,494	Large-scale dataset for anatomy segmentation in nuclear imaging.	Link
JSRT	Chest X-Ray	247	Dataset of chest radiographs for lung segmentation.	Link
PAX-Ray	Chest X-Ray	852	Dataset for fine-grained anatomy segmentation.	Link
PAX-Ray++	Chest X-Ray	14,754	Enhanced version of PAX-Ray with additional annotations.	Link
DUKE	Optical Coherence Tomography	110	OCT dataset for retinal layer segmentation.	Link
RAVIR	Infrared Reflectance Imaging	23	Dataset for retinal vessel segmentation from infrared images.	Link
Teeth	Panoramic X-Ray	598	Dataset of dental X-rays for teeth segmentation.	Link

---

📁 Repository Structure

Here’s an overview of the repository's structure:

src/
│
├── data/                        # PyTorch dataloaders for all datasets
│
├── prepare_data/
│   ├── prepare_bs80k/                   # Scripts for processing the anatomical segmentations of BS80k dataset
│   ├── prepare_jsrt/                    # Scripts for processing the JSRT dataset
│   ├── prepare_paxray/                 # Scripts for processing the PAX-Ray dataset
│   ├── prepare_paxraypp/               # Scripts for processing the PAX-Ray++ dataset
│   ├── prepare_duke/                    # Scripts for processing the DUKE dataset
│   ├── prepare_ravir/                   # Scripts for processing the RAVIR dataset
│   └── prepare_teeth_kaggle/                   # Scripts for processing the Teeth dataset
│
├── visualization/
│   └── visualize.py             # Utility scripts for visualizing dataset samples
│
├── Notebooks/
│   └── Download_datasets.ipynb    # shows how to setup all the datasets
│   └── Dataloader_example.ipynb   # Examples for to setup the dataloaders for each dataset and visualize annotations

---

⚙️ Features

🛠 Data Preparation

Each dataset includes standardized processing scripts to ensure consistency across datasets. These scripts:

• Download the raw data.
• Normalize and preprocess the images.
• Prepare data splits (train/val/test).
• Create PyTorch-friendly formats.

📊 Data Visualization

Use the scripts in src/visualization to visualize the datasets, ground-truth labels, and predictions for easy inspection of data quality.

🚀 Training Pipeline Integration

Prepared datasets can be plugged directly into any PyTorch-based segmentation model for training and evaluation.

---

🚀 Getting Started

1️⃣ Clone the Repository

git clone https://github.com/yourusername/2D-Anatomy-Segmentation-Datasets.git
cd 2D-Anatomy-Segmentation-Datasets

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Prepare the Data

Run the dataset preparation script for the dataset of interest. For example:

sh src/prepare_data/prepare_bs80k/get_bs80k_full.sh

sh src/prepare_data/prepare_jsrt/get_jsrt_full.sh

...

4️⃣ Visualize the Data

To visualize samples:

python src/visualization/visualize.py --dataset BS80k

---

🤝 Contributions

Contributions are welcome! If you'd like to add support for more datasets, improve baseline performance, or enhance data preparation scripts, feel free to open a pull request.

---

📜 License

This project is licensed under the MIT License.

---

💬 Contact

For questions or feedback, feel free to reach out:
📧 Email: constantinseibold[at]gmail.com
🔗 GitHub Issues

---

Enjoy building robust segmentation models with these datasets! 😊