MI inverse inference

Code for the inverse inference of infarcted area from ECG and MRI for myocardial infarction (MI) patients. This is achieved within a cardiac digital twin (CDT) framework, where the anatomical twinning personalizes the geometrical model, while functional twinning personalizes the electrophysiological model.

Notes:

1. Cardiac_Personalisation-SenAnalysis fold only contain partial code for the ECG simulation and the sensitivity analysis. For the full Eikonal-based ECG simulation code, please contact Dr Julia Camps.
2. Cobiveco fold only contrain partial code for converting biventicle mesh into cobiveco mesh. For the complete Cobiveco mesh reconstruction code, please visit KIT-IBE Cobiveco Github repository.

Package dependencies:

This repository is based on PyTorch, running on a computer with 3.50~GHz Intel(R) Xeon(R) E-2146G CPU and an NVIDIA GeForce RTX 3060.

Dataset:

Our network is trained based on UKB dataset, which contains paired multi-view MRIs and ECG data. We have reconstruced the cardiac geometry from the multi-view MRIs and converted it into cobiveco mesh. The scar/border zone area were assigned on the mesh and subsequently used for ECG simulation of MI patients.

The data folder should be like:

tree
`-- UKB_clinical_data
    `-- patientID
    |    |-- patientID_cobiveco_AHA17.vtu
    |    |-- patientID_heart_cobiveco.vtu
    |    |-- patientID_simulated_ECG_xxx_subendo.csv
    |    |-- patientID_simulated_ECG_xxx_transmural.csv
    |    |-- ...
    |    |-- patientID_electrodePosition.csv
		

Citation:

If you find this code useful in your research, please consider citing:

@article{jounral/TMI/li2024,
  title={Towards Enabling Cardiac Digital Twins of Myocardial Infarction Using Deep Computational Models for Inverse Inference},
  author={Li, Lei and Camps, Julia and Wang, Zhinuo and Banerjee, Abhirup and Rodriguez, Blanca and Grau, Vicente},
  journal={IEEE Transactions on Medical Imaging},
  year={2024}
}