/graph_ECGI

Repository for Geometric Model (Jiang et al, IEEE TMI 2023 / MICCAI 2020), Hybrid SSM (Jiang et al, IEEE TMI 2024 / MICCAI 2021), and Personalized Neural Surrogate (ICLR 2023 / MICCAI 2022).

Primary LanguagePython

Running guidance

  1. Check your cuda version. I've tested cu117 and it is able to work.
  2. Python version >= 3.8 would be recommended.
  3. To install the python packages, change the variable CUDA in the script req_torch_geo.sh, then run it.
  4. Model list:
    1. demo01: ST-GCNN heart-in-heart-out
    2. demo02: Euclidean (Sandesh et al, IPMI 2019)
    3. demo03: meta-model (Jiang et al, ICLR 2023)
    4. demo04: ST-GCNN torso-in-heart-out (Jiang et al, IEEE TMI 2022 and MICCAI 2020)
    5. demo05: meta-model with neural ODE
    6. demo06: meta-model with initial condition masks (Jiang et al, MICCAI 2022)
    7. demo07: HybridSSM unsupervised learning (Jiang et al, IEEE TMI 2024 and MICCAI 2021)
    8. demo08 HybridSSM mixed learning (Jiang et al, IEEE TMI 2024)
  5. To train the model, run the following command: 
    python main.py --config demo01 --stage 1
  6. To evaluate the model, run the following command: 
    python main.py --config demo01 --stage 2
  7. To perform meta-evaluation on the model, run the following command: 
    python main.py --config demo03 --stage 3
  8. To make graphs based on the pre-defined geometry, run the following command: 
    python main.py --config demo01 --stage 0
  9. To save evaluation results for visualization, add --tag <file name> to the end of evaluation command.

Citation

Please cite the following if you use the data or the model in your work:

  1. Geometric Model
@ARTICLE{9932432,
  author={Jiang, Xiajun and Toloubidokhti, Maryam and Bergquist, Jake and Zenger, Brian and Good, Wilson W. and MacLeod, Rob S. and Wang, Linwei},
  journal={IEEE Transactions on Medical Imaging}, 
  title={Improving Generalization by Learning Geometry-Dependent and Physics-Based Reconstruction of Image Sequences}, 
  year={2023},
  volume={42},
  number={2},
  pages={403-415},
  keywords={Geometry;Heart;Image reconstruction;Physics;Training data;Imaging;Torso;Geometric deep learning;inverse problems;physics-based deep learning},
  doi={10.1109/TMI.2022.3218170}}

@inproceedings{jiang2020learning,
  title={Learning geometry-dependent and physics-based inverse image reconstruction},
  author={Jiang, Xiajun and Ghimire, Sandesh and Dhamala, Jwala and Li, Zhiyuan and Gyawali, Prashnna Kumar and Wang, Linwei},
  booktitle={Medical Image Computing and Computer Assisted Intervention--MICCAI 2020: 23rd International Conference, Lima, Peru, October 4--8, 2020, Proceedings, Part VI 23},
  pages={487--496},
  year={2020},
  organization={Springer}
}
  1. Hybrid SSM
@ARTICLE{10471622,
  author={Jiang, Xiajun and Missel, Ryan and Toloubidokhti, Maryam and Gillette, Karli and Prassl, Anton J. and Plank, Gernot and Milan Horáček, B. and Sapp, John L. and Wang, Linwei},
  journal={IEEE Transactions on Medical Imaging}, 
  title={Hybrid Neural State-Space Modeling for Supervised and Unsupervised Electrocardiographic Imaging}, 
  year={2024},
  volume={},
  number={},
  pages={1-1},
  keywords={Electrocardiography;Bayes methods;Heart;Filtering;Data models;Image reconstruction;Physics;Image reconstruction;Bayesian filter;Neural ODE;Graph convolution},
  doi={10.1109/TMI.2024.3377094}}

@inproceedings{jiang2021label,
  title={Label-free physics-informed image sequence reconstruction with disentangled spatial-temporal modeling},
  author={Jiang, Xiajun and Missel, Ryan and Toloubidokhti, Maryam and Li, Zhiyuan and Gharbia, Omar and Sapp, John L and Wang, Linwei},
  booktitle={Medical Image Computing and Computer Assisted Intervention--MICCAI 2021: 24th International Conference, Strasbourg, France, September 27--October 1, 2021, Proceedings, Part VI 24},
  pages={361--371},
  year={2021},
  organization={Springer}
}
  1. Personalized Neural Surrogate
@inproceedings{
jiang2023sequential,
title={Sequential Latent Variable Models for Few-Shot High-Dimensional Time-Series Forecasting},
author={Xiajun Jiang and Ryan Missel and Zhiyuan Li and Linwei Wang},
booktitle={The Eleventh International Conference on Learning Representations },
year={2023},
url={https://openreview.net/forum?id=7C9aRX2nBf2}
}

@inproceedings{jiang2022few,
  title={Few-Shot Generation of Personalized Neural Surrogates for Cardiac Simulation via Bayesian Meta-learning},
  author={Jiang, Xiajun and Li, Zhiyuan and Missel, Ryan and Zaman, Md Shakil and Zenger, Brian and Good, Wilson W and MacLeod, Rob S and Sapp, John L and Wang, Linwei},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={46--56},
  year={2022},
  organization={Springer}
}