<p align=center>Vision foundation model for 3D magnetic resonance imaging segmentation, classification, and registration</p>

This paper has been accepted by Medical Image Analysis.

• Our paper: https://www.sciencedirect.com/science/article/pii/S1361841526000617

Quick Start

This repository provides a minimal example in QuickStart.py for loading released encoder/backbone weights.

Available weights in this directory:

• Triad-PlainConvUNet-MAE.pth https://drive.google.com/file/d/1Mc5owEFhWkroe5Hjnk7Ex8v6Z-aBKF3U/view?usp=drive_link

• Triad-PlainConvUNet-SimMIM.pth https://drive.google.com/file/d/1EkrYbuNI64yi1_Yl4JZUM5yZ5K0mHzR3/view?usp=drive_link

• Triad-SwinB-MAE.pth https://drive.google.com/file/d/1F_6TNrCxPyqk-bPzXj0HHLxFy9mbNzjl/view?usp=drive_link

• Triad-SwinB-SimMIM.pth https://drive.google.com/file/d/1icLjmSpTdEAA9kEW3BWHnAYv-hsXMYxS/view?usp=drive_link

PlainConvUNet

The default runnable example in QuickStart.py loads:

ckpt = torch.load("Triad-PlainConvUNet-MAE.pth", weights_only=False)

You can switch it to:

• Triad-PlainConvUNet-MAE.pth
• Triad-PlainConvUNet-SimMIM.pth

Then run:

python QuickStart.py

Swin-B

At the bottom of QuickStart.py, a Swin example is provided as commented code.

Uncomment the Swin block, then choose one checkpoint:

• Triad-SwinB-MAE.pth
• Triad-SwinB-SimMIM.pth

Set:

ckpt = torch.load("Triad-SwinB-SimMIM.pth", weights_only=False)

Then run:

python QuickStart.py

Citation

@article{WANG2026103992,
title = {Vision foundation model for 3D magnetic resonance imaging segmentation, classification, and registration},
journal = {Medical Image Analysis},
volume = {110},
pages = {103992},
year = {2026},
issn = {1361-8415},
doi = {https://doi.org/10.1016/j.media.2026.103992},
url = {https://www.sciencedirect.com/science/article/pii/S1361841526000617},
author = {Shansong Wang and Mojtaba Safari and Qiang Li and Chih-Wei Chang and Richard {LJ Qiu} and Justin Roper and David S. Yu and Xiaofeng Yang},
}

Acknowledgments

• This project is based on VoCo v2: https://github.com/Luffy03/Large-Scale-Medical