<img src="ICON_universal_model.png" width="35" height="35"> uniGradICON: A Foundation Model for Medical Image Registration

<img src="https://github.com/uncbiag/unigradicon/actions/workflows/test_readme_works.yml/badge.svg">

News

• 🔥 uniGradICON Slicer Extension is available in 3D Slicer 5.7.0 Extensions Manager (12/2024)
• 🏆 multiGradICON wins the best oral presentation at 2024 MICCAI Workshop for Biomedical Image Registration (WBIR) (10/2024)
• 🔥 uniGradICON has been used as a baseline in LUMIR Brain MRI Registration Challenge (6/2024)

Introduction

uniGradICON is based on GradICON but trained on several different datasets (see details below). The result is a deep-learning-based registration model that works well across datasets. More results can be found here. To get you started quickly there is also an easy to use Slicer extension.

<div align="center"> <img src="IntroFigure.jpg" width=550 alt="teaser"> </div>

This is the repository for the following papers

uniGradICON: A Foundation Model for Medical Image Registration
Tian, Lin and Greer, Hastings and Kwitt, Roland and Vialard, Francois-Xavier and Estepar, Raul San Jose and Bouix, Sylvain and Rushmore, Richard and Niethammer, Marc
MICCAI 2024 https://arxiv.org/abs/2403.05780

multiGradICON: A Foundation Model for Multimodal Medical Image Registration
Demir, Basar and Tian, Lin and Greer, Thomas Hastings and Kwitt, Roland and Vialard, Francois-Xavier and Estepar, Raul San Jose and Bouix, Sylvain and Rushmore, Richard Jarrett and Ebrahim, Ebrahim and Niethammer, Marc
MICCAI Workshop on Biomedical Image Registration (WBIR) 2024 https://arxiv.org/abs/2408.00221

Easy to use and install

The pre-trained uniGradICON and multiGradICON can be used via CLI, colab notebook, and Slicer Extension. The model weights will be downloaded automatically. You can also find the model weights here.

👉 Inference via CLI

Installation

python3 -m venv unigradicon_virtualenv
source unigradicon_virtualenv/bin/activate

pip install unigradicon

To register one pair of image

wget https://www.hgreer.com/assets/slicer_mirror/RegLib_C01_1.nrrd
wget https://www.hgreer.com/assets/slicer_mirror/RegLib_C01_2.nrrd

unigradicon-register --fixed=RegLib_C01_2.nrrd --fixed_modality=mri --moving=RegLib_C01_1.nrrd --moving_modality=mri --transform_out=trans.hdf5 --warped_moving_out=warped_C01_1.nrrd

To register without instance optimization (IO)

unigradicon-register --fixed=RegLib_C01_2.nrrd --fixed_modality=mri --moving=RegLib_C01_1.nrrd --moving_modality=mri --transform_out=trans.hdf5 --warped_moving_out=warped_C01_1.nrrd --io_iterations None

To use a different similarity measure in the IO. We currently support three similarity measures

• LNCC: lncc
• Squared LNCC: lncc2
• MIND SSC: mind

unigradicon-register --fixed=RegLib_C01_2.nrrd --fixed_modality=mri --moving=RegLib_C01_1.nrrd --moving_modality=mri --transform_out=trans.hdf5 --warped_moving_out=warped_C01_1.nrrd --io_iterations 50 --io_sim lncc2

To load specific model weight in the inference. We currently support uniGradICON and multiGradICON.

unigradicon-register --fixed=RegLib_C01_2.nrrd --fixed_modality=mri --moving=RegLib_C01_1.nrrd --moving_modality=mri --transform_out=trans.hdf5 --warped_moving_out=warped_C01_1.nrrd --model multigradicon

To mask out the background using the provided segmentation before registration (segmentations for both moving and fixed images are necessary for accurate registration):

unigradicon-register --fixed=RegLib_C01_2.nrrd --fixed_modality=mri --fixed_segmentation=[fixed_image_segmentation_file_name] --moving=RegLib_C01_1.nrrd --moving_modality=mri --moving_segmentation=[moving_image_segmentation_file_name] --transform_out=trans.hdf5 --warped_moving_out=warped_C01_1.nrrd --io_iterations None

To apply loss function masking using the provided segmentations in the IO:

unigradicon-register --fixed=RegLib_C01_2.nrrd --fixed_modality=mri --fixed_segmentation=[fixed_image_segmentation_file_name] --moving=RegLib_C01_1.nrrd --moving_modality=mri --moving_segmentation=[moving_image_segmentation_file_name] --transform_out=trans.hdf5 --warped_moving_out=warped_C01_1.nrrd --io_iterations 50 --io_sim lncc2 --loss_function_masking

This loss function ensures proper intensity adjustments for registration tasks requiring mass conservation, utilizing the change of variables rule from integration. To be effective, images must be in an intensity space where conservation holds. This loss function is specifically valid for CT modality, where -1000 HU represents air. To apply determinant-based intensity correction during registration in the IO:

unigradicon-register --fixed=RegLib_C01_2.nrrd --fixed_modality=mri --fixed_segmentation=[fixed_image_segmentation_file_name] --moving=RegLib_C01_1.nrrd --moving_modality=mri --moving_segmentation=[moving_image_segmentation_file_name] --transform_out=trans.hdf5 --warped_moving_out=warped_C01_1.nrrd --io_iterations 50 --io_sim lncc2 --intensity_conservation_loss

To warp an image

unigradicon-warp --fixed [fixed_image_file_name] --moving [moving_image_file_name]  --transform trans.hdf5 --warped_moving_out warped.nii.gz --linear

To warp a label map

unigradicon-warp --fixed [fixed_image_file_name] --moving [moving_image_segmentation_file_name]  --transform trans.hdf5 --warped_moving_out warped_seg.nii.gz --nearest_neighbor

👉 Inference via colab notebook

We provide a colab notebook where the users can directly access and visualize the output of uniGradICON network.

👉 Inference via Slicer Extension

A Slicer extensions is available here. It is an official Slicer Extension and can be installed via the Slicer Extension Manager. This requires Slicer >=5.7.0. Please make sure to install the Slicer PyTorch extension before as uniGradICON depends on it.

Training and testing data

uniGradICON has currently been trained and tested on the following datasets.

Training data:

<table> <tr> <td> </td> <td>Dataset</td> <td>Anatomical region</td> <td># of patients</td> <td># per patient</td> <td># of pairs</td> <td>Type</td> <td>Modality</td> </tr> <tr> <td>1.</td> <td>COPDGene</td> <td>Lung</td> <td>899</td> <td>2</td> <td>899</td> <td>Intra-pat.</td> <td>CT</td> </tr> <tr> <td>2.</td> <td>OAI</td> <td>Knee</td> <td>2532</td> <td>1</td> <td>3,205,512</td> <td>Inter-pat.</td> <td>MRI</td> </tr> <tr> <td>3.</td> <td>HCP</td> <td>Brain</td> <td>1076</td> <td>1</td> <td>578,888</td> <td>Inter-pat.</td> <td>MRI</td> </tr> <tr> <td>4.</td> <td>L2R-Abdomen</td> <td>Abdomen</td> <td>30</td> <td>1</td> <td>450</td> <td>Inter-pat.</td> <td>CT</td> </tr> </table>

Testing data:

<table> <tr> <td> </td> <td>Dataset</td> <td>Anatomical region</td> <td># of patients</td> <td># per patient</td> <td># of pairs</td> <td>Type</td> <td>Modality</td> </tr> <tr> <td>5.</td> <td>Dirlab-COPDGene</td> <td>Lung</td> <td>10</td> <td>2</td> <td>10</td> <td>Intra-pat.</td> <td>CT</td> </tr> <tr> <td>6.</td> <td>OAI-test</td> <td>Knee</td> <td>301</td> <td>1</td> <td>301</td> <td>Inter-pat.</td> <td>MRI</td> </tr> <tr> <td>7.</td> <td>HCP-test</td> <td>Brain</td> <td>32</td> <td>1</td> <td>100</td> <td>Inter-pat.</td> <td>MRI</td> </tr> <tr> <td>8.</td> <td>L2R-NLST-val</td> <td>Lung</td> <td>10</td> <td>2</td> <td>10</td> <td>Intra-pat.</td> <td>CT</td> </tr> <tr> <td>9.</td> <td>L2R-OASIS-val</td> <td>Brain</td> <td>20</td> <td>1</td> <td>19</td> <td>Inter-pat.</td> <td>MRI</td> </tr> <tr> <td>10.</td> <td>IXI-test</td> <td>Brain</td> <td>115</td> <td>1</td> <td>115</td> <td>Atlas-pat.</td> <td>MRI</td> </tr> <tr> <td>11.</td> <td>L2R-CBCT-val</td> <td>Lung</td> <td>3</td> <td>3</td> <td>6</td> <td>Intra-pat.</td> <td>CT/CBCT</td> </tr> <tr> <td>12.</td> <td>L2R-CTMR-val</td> <td>Abdomen</td> <td>3</td> <td>2</td> <td>3</td> <td>Intra-pat.</td> <td>CT/MRI</td> </tr> <tr> <td>13.</td> <td>L2R-CBCT-train</td> <td>Lung</td> <td>3</td> <td>11</td> <td>22</td> <td>Intra-pat.</td> <td>CT/CBCT</td> </tr> </table>

Get involved

Our goal is to continuously improve the uniGradICON model, e.g., by training on more datasets with additional diversity.