datasets:
- AnonRes/OpenMind
license: cc-by-4.0
pipeline_tag: image-segmentation
tags:
- medical
OpenMind Benchmark 3D SSL Models
This repository hosts pre-trained checkpoints from the OpenMind benchmark, featuring architectures introduced in the paper Primus: Enforcing Attention Usage for 3D Medical Image Segmentation.
Primus Paper: Primus: Enforcing Attention Usage for 3D Medical Image Segmentation
OpenMind Paper: An OpenMind for 3D medical vision self-supervised learning
Pre-training codebase: MIC-DKFZ/nnssl
Downstream framework: MIC-DKFZ/nnUNet
Dataset: AnonRes/OpenMind
Overview
This repository hosts pre-trained checkpoints from the OpenMind benchmark: the first extensive benchmark study for self-supervised learning (SSL) on 3D medical imaging data.
Each model was pre-trained using a particular SSL method on the OpenMind Dataset, a large-scale, standardized collection of public brain MRI datasets.
Primus and PrimusV2 are Transformer-centric segmentation architectures designed to overcome the over-reliance on convolutional blocks. Through high-resolution tokens and iterative patch embeddings, these models achieve state-of-the-art results in 3D medical image segmentation.
These models are intended for downstream fine-tuning. We recommend using the fine-tuning frameworks for segmentation and classification available in the nnU-Net repository or the adaptation repository.
Model Variants
We release SSL checkpoints for two backbone architectures:
- ResEnc-L: A CNN-based encoder [a, b]
- Primus-M: A transformer-based encoder [Primus paper]
Each encoder has been pre-trained using one of the following SSL techniques:
| Method | Description |
|---|---|
| Volume Contrastive (VoCo) | Contrastive pretraining method for 3D volumes |
| VolumeFusion (VF) | Spatial volume fusion-based segmentation SSL method |
| Models Genesis (MG) | Reconstruction and denoising based pretraining method |
| Masked Autoencoders (MAE) | Default reconstruction based pretraining method |
| Spark 3D (S3D) | Sparse reconstruction based pretraining method (CNN only) |
| SimMIM | Simple masked reconstruction based pretraining method (TR only) |
| SwinUNETR SSL | Rotation, Contrastive and Reconstruction based pre-training method. |
| SimCLR | Transfer of 2D Contrastive learning baseline method to 3D |
Citation
If you use these models or the Primus architecture, please cite:
@article{wald2025primus,
title={Primus: Enforcing Attention Usage for 3D Medical Image Segmentation},
author={Wald, Tassilo and Roy, Saikat and Isensee, Fabian and Ulrich, Constantin and Ziegler, Sebastian and Trofimova, Dasha and Stock, Raphael and Baumgartner, Michael and K\u00f6hler, Gregor and Maier-Hein, Klaus},
journal={arXiv preprint arXiv:2503.01835},
year={2025}
}
@article{wald2024openmind,
title={An OpenMind for 3D medical vision self-supervised learning},
author={Wald, Tassilo and Ulrich, Constantin and Suprijadi, Jue and Ziegler, Sebastian and Nohel, Michal and Peretzke, Richard and Isensee, Fabian and Maier-Hein, Klaus H},
journal={arXiv preprint arXiv:2412.17041},
year={2024}
}