MedOpenSeg

Open-World Medical Segmentation with Memory-Augmented Transformers 🧠🧪

This repository contains the code to reproduce the experiments reported in:
“MedOpenSeg: Open-World Medical Segmentation with Memory-Augmented Transformers.”
Accepted at BMVC 2025 🎉

---

📦 Quick Installation

# Install PyTorch (recommended version: 2.1.0, adjust CUDA as needed)
pip install torch==2.1.0 torchvision==0.16.0 --index-url https://download.pytorch.org/whl/cu121

# Minimal extra packages
pip install numpy scipy nibabel SimpleITK tqdm monai

---

📂 Datasets

Experiments are conducted on:

• AMOS 2022 (multi-organ CT, 15 abdominal organs)
• BTCV (Synapse) (multi-organ abdominal CT)
• MSD – Pancreas (pancreas + tumors)

---

🚀 Usage

Training

python trainV.py --config configs/amos.yaml --data_root DATA/AMOS --exp_name amos_os

Inference (masks + novelty maps)

python inference_V.py   --config configs/amos.yaml   --checkpoint runs/amos_os/best.ckpt   --data_root DATA/AMOS   --save_dir outputs/amos_os

---

📖 How to cite

If this work helps your research, please cite:

---

🙏 Thank you for your interest in MedOpenSeg!