Home

Awesome

<h1 align="center">ā— Medical SAM 2: Segment Medical Images As Video Via Segment Anything Model 2</h1> <p align="center"> <a href="https://discord.gg/DN4rvk95CC"> <img alt="Discord" src="https://img.shields.io/discord/1146610656779440188?logo=discord&style=flat&logoColor=white"/></a> <img src="https://img.shields.io/static/v1?label=license&message=GPL&color=white&style=flat" alt="License"/> </p>

Medical SAM 2, or say MedSAM-2, is an advanced segmentation model that utilizes the SAM 2 framework to address both 2D and 3D medical image segmentation tasks. This method is elaborated on the paper Medical SAM 2: Segment Medical Images As Video Via Segment Anything Model 2.

šŸ”„ A Quick Overview

<div align="center"><img width="880" height="350" src="https://github.com/MedicineToken/Medical-SAM2/blob/main/vis/framework.png"></div>

šŸ©» 3D Abdomen Segmentation Visualisation

<div align="center"><img width="420" height="420" src="https://github.com/MedicineToken/Medical-SAM2/blob/main/vis/example.gif"></div>

šŸ§ Requirement

Install the environment:

conda env create -f environment.yml

conda activate medsam2

You can download SAM2 checkpoint from checkpoints folder:

bash download_ckpts.sh

Further Note: We tested on the following system environment and you may have to handle some issue due to system difference.

Operating System: Ubuntu 22.04
Conda Version: 23.7.4
Python Version: 3.12.4

We released our pretrain weight here

šŸŽÆ Example Cases

Download REFUGE or BCTV or your own dataset and put in the data folder, create the folder if it does not exist āš’ļø

2D case - REFUGE Optic-cup Segmentation from Fundus Images

Step1: Download pre-processed REFUGE dataset manually from here, or using command lines:

wget https://huggingface.co/datasets/jiayuanz3/REFUGE/resolve/main/REFUGE.zip

unzip REFUGE.zip

Step2: Run the training and validation by:

python train_2d.py -net sam2 -exp_name REFUGE_MedSAM2 -vis 1 -sam_ckpt ./checkpoints/sam2_hiera_small.pt -sam_config sam2_hiera_s -image_size 1024 -out_size 1024 -b 4 -val_freq 1 -dataset REFUGE -data_path ./data/REFUGE

3D case - Abdominal Multiple Organs Segmentation

Step1: Download pre-processed BTCV dataset manually from here, or using command lines:

wget https://huggingface.co/datasets/jiayuanz3/btcv/resolve/main/btcv.zip

unzip btcv.zip

Step2: Run the training and validation by:

python train_3d.py -net sam2 -exp_name BTCV_MedSAM2 -sam_ckpt ./checkpoints/sam2_hiera_small.pt -sam_config sam2_hiera_s -image_size 1024 -val_freq 1 -prompt bbox -prompt_freq 2 -dataset btcv -data_path ./data/btcv

šŸšØ News

šŸ“ Cite

@misc{zhu_medical_2024,
   title={Medical SAM 2: Segment medical images as video via Segment Anything Model 2},
   author={Jiayuan Zhu and Yunli Qi and Junde Wu},
   year = {2024},
   eprint={2408.00874},
   archivePrefix={arXiv},
   primaryClass={cs.CV}
}