Robust deformable image registration using synthetic data and transfer learning

Iris D. Kolenbrander, Matteo Maspero, Josien P.W. Pluim

This repository contains code for pre-training and fine-tuning deep learning-based registration models. It also provides the download links to the pre-trained model, which can be used for fine-tuning on target datasets. Paper is available here.

Citation

If you use this repository or its contents in your research, please cite the following paper:

Kolenbrander, I. D., Maspero, M., & Pluim, J. P. W. (2025). Robust deformable image registration using synthetic data and transfer learning. Machine Learning for Biomedical Imaging, 3(August 2025 issue), 287–316.

---

Models

The trained model weights are available for download.

Download model weights

Data

The synthetic data labels and statistical model for data augmentation (for Lung CT) are available for download. In addition, to help users test the pipeline, we provide toy data (only 5/2 train/val cases).

Download data

---

Code

We used PyTorch Lightning for training and inference. .
The code is structured into two main stages:

python -m configs.stage_01_pretrain_synthetic_data


### 1. Pre-training on synthetic data
python -m configs.stage_01_pretrain_synthetic_data

### 2. Training on target datasets (e.g., lung CT)
#### 2.1. From-scratch training
python -m configs.stage_02_finetune_lung_CT_data

#### 2.2. Fine-tuning (after pre-training on synthetic data)
python -m configs.stage_02_finetune_lung_CT_data --load_model pretrained_model