🔬 IMPACT: A Generic Semantic Loss for Multimodal Image Registration

IMPACT is a novel, task-agnostic similarity metric designed for multimodal medical image registration. Instead of relying on intensity based metric, handcrafted descriptors or training task-specific models, IMPACT reuses powerful segmentation foundation models (e.g., TotalSegmentator, SAM) as generic feature extractors. These deep features are used to define a semantic similarity loss, optimized directly in registration frameworks like Elastix or VoxelMorph.

📚 Reference

🔗 IMPACT: A Generic Semantic Loss for Multimodal Image Registration Valentin Boussot, Cédric Hémon, Jean-Claude Nunes, Jason Dowling, Simon Rouzé, Caroline Lafond, Anaïs Barateau, Jean-Louis Dillenseger arXiv:2503.24121 – Under review

---

✨ Key Features

• Generic, Training-free
  No need for task-specific training, IMPACT reuses powerful representations from large-scale pretrained segmentation models.

• Flexible model integration
  Compatible with TorchScript 2D/3D models (e.g., TotalSegmentator, SAM2.1, MIND), supporting multi-layer and multi-model fusion, multi-resolution setups, and fully open to experimentation with custom architectures and configurations.

• Jacobian vs Static optimization modes
  Choose between fully differentiable Jacobian mode (for downsampling models) and fast inference-only Static mode, depending on model type and computation time constraints.

• Robust across modalities
  Handles complex multimodal scenarios (CT/CBCT, MR/CT) using a unified semantic loss robust to intensity variations.

• Benchmark-proven Ranked in the top participants of multiple Learn2Reg challenges, showing state-of-the-art performance across diverse tasks (thorax, abdomen, pelvis, CT/CBCT/MRI).

• Seamless integration with Elastix
  Natively implemented as a standard Elastix metric, IMPACT inherits all the strengths of classical registration: multi-resolution strategies, mask support, sparce deformation models, and full reproducibility. It also handles images of different sizes, resolutions, and fields of view, making it ideal for real-world clinical datasets with heterogeneous inputs.

• Efficient runtime for standard registration tasks

  • ~150 seconds (Static mode)
  • ~300 seconds (Jacobian mode)

• Docker-ready for quick deployment
  Run out of the box with a single Docker command, no need to install dependencies manually.

---

🏆 Challenge Results

IMPACT has demonstrated strong generalization performance across multiple tasks without training. 🔗 Learn2Reg Challenge

Challenge	Task	Rank
Learn2Reg 2021	CT Lung Registration	🥉 3rd
Learn2Reg 2023	Thorax CBCT	🥉 Top-6
Learn2Reg 2023	Abdomen MR→CT	🥈 2nd

---

🔬 Modular and Extensible Design

Beyond its performance, IMPACT is designed as a modular platform that facilitates systematic experimentation with pretrained models, feature layers, and distance functions. This flexibility enables researchers to explore various feature extraction methods, fostering innovation and adaptability in multimodal image registration tasks.

---

🚀 Quick Start with Docker

You can quickly test the IMPACT metric using the provided Docker environment:

git clone https://github.com/vboussot/ImpactLoss.git
cd ImpactLoss

Build the Docker image

docker build -t elastix_impact Docker

Then, run Elastix with your own data:

docker run --rm --gpus all \
  -v "./Data:/Data" \
  -v "./Out:/Out" \
  elastix_impact

Make sure that the Data/ folder contains:

• FixedImage.mha, MovingImage.mha your input images to be registered. These files can be provided in either .mha or .nii.gz format.
• ParameterMap.txt using Impact configuration. 👉 See ParameterMaps/README.md for detailed configuration examples.
• A Data/Models/ directory with TorchScript models. 👉 See Data/Models/README.md for model download instructions.

See Docker/README.md for full details and usage examples.

💡 You can also directly pull the prebuilt image from Docker Hub:

docker pull vboussot/elastix_impact

👉

---

🛠️ Manual Build Instructions (without Docker)

Build Elastix with IMPACT support directly on your machine.

📦 (Optional) Get LibTorch (if not already installed)

Download and extract the C++ distribution of LibTorch (with or without CUDA) from the official website:

🔗 https://pytorch.org/

📦 (Optional) Build ITK (if not already installed):

git clone https://github.com/InsightSoftwareConsortium/ITK.git
mkdir ITK-build ITK-install
cd ITK-build
cmake -DCMAKE_INSTALL_PREFIX=../ITK-install ../ITK
make install
cd ..

🧱 Build Instructions

1. Clone the ImpactElastix repository:

git clone https://github.com/vboussot/ImpactElastix.git

2. Create build and install directories:

mkdir ImpactElastix-build ImpactElastix-install
cd ImpactElastix-build

3. Configure the build with CMake:

cmake -DTorch_DIR=../libtorch/share/cmake/Torch/ \
      -DITK_DIR=../ITK-install/lib/cmake/ITK-6.0/ \
      -DCMAKE_INSTALL_PREFIX=../ImpactElastix-install \
      ../ImpactElastix

• Torch_DIR: path to the CMake config directory of LibTorch (usually inside libtorch/share/cmake/Torch/)
• ITK_DIR: path to the CMake config directory of ITK, typically inside your ITK install folder (e.g., ITK-install/lib/cmake/ITK-*)

4. Build and install Elastix with IMPACT:

make install

The final binaries will be located in:

../ImpactElastix-install/bin/elastix

Before running elastix, make sure the required shared libraries are accessible at runtime by setting the LD_LIBRARY_PATH:

export LD_LIBRARY_PATH=lib/libtorch/lib:$LD_LIBRARY_PATH
export LD_LIBRARY_PATH=ImpactElastix-install/lib:$LD_LIBRARY_PATH

You can then run:

../ImpactElastix-install/bin/elastix

⚙️ Run Elastix

To use IMPACT, start by downloading the pretrained TorchScript models.
👉 See Data/Models/README.md for download instructions.

Elastix is executed as usual, using a parameter map configured to use the IMPACT metric.
👉 Refer to ParameterMaps/README.md for detailed configuration examples.

⚠️ Preprocessing Recommendation
Input images must be preprocessed consistently with the training of the selected model.
For TotalSegmentator-based models, images should be in canonical orientation.

Apply the appropriate preprocessing depending on the model:

• ImageNet-based models (e.g., SAM2.1, DINOv2):

  • Normalize intensities to [0, 1]
  • Then standardize with mean 0.485 and standard deviation 0.229

• MRI models (e.g., TS/M730–M733):

  • Standardize intensities to zero mean and unit variance

• CT models (e.g., all other TotalSegmentator variants, MIND):

  • Clip intensities to [-1024, 276] HU
  • Then normalize by centering at -370 HU and scaling by 436.6

A complete example of how to run registration with IMPACT is provided in:
👉 run_impact_example.py