dataset/data_preprocessing.py file does data preprocessing tasks, processed the data in defined root, and returns a dict to pair the information by timestamp, with force, image, and pose information.
dataset/pairing.py we can get the paired image with force and pose by linear assignment from timestamp.
raft/generate_gt_stepwise.py will generate gt flow between 2 frames. When the deformation is too large, we will use an intermediate frame to do a step-wise flow estimation.
The code implementation of the network see networks/c2f_stiff_second_network.py
python train.py --opt options/train_c2f_stiff_second.yaml
The work has been Published in Medical Image Analysis
@article{jiang2023defcor,
title={DefCor-Net: Physics-aware ultrasound deformation correction},
author={Jiang, Zhongliang and Zhou, Yue and Cao, Dongliang and Navab, Nassir},
journal={Medical Image Analysis},
pages={102923},
year={2023},
publisher={Elsevier}
}
A previous study focused on ultrasound image correction was published in RAL 2021
@article{jiang2021deformation,
title={Deformation-aware robotic 3D ultrasound},
author={Jiang, Zhongliang and Zhou, Yue and Bi, Yuan and Zhou, Mingchuan and Wendler, Thomas and Navab, Nassir},
journal={IEEE Robotics and Automation Letters},
volume={6},
number={4},
pages={7675--7682},
year={2021},
publisher={IEEE}
}