ECGFlowCMR

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

★ 10 Open GitHub ↗

[KDD 2026] ECGFlowCMR: Pretraining with ECG-Generated Cine CMR Improves Cardiac Disease Classification and Phenotype Prediction

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1> ECGFlowCMR: Pretraining with ECG-Generated Cine CMR Improves Cardiac Disease Classification and Phenotype Prediction </h1>

<h5 align="center"> If you find this project useful, please give us a star🌟.

<a href='https://arxiv.org/pdf/2601.20904'><img src='https://img.shields.io/badge/Paper-Arxiv-red'></a>
<a href='https://huggingface.co/XiaochengFang/ECGFlowCMR/tree/main'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Models-blue'>
<a href='https://www.ukbiobank.ac.uk/'><img src='https://img.shields.io/badge/Dataset-Huggingface-yellow'>

<p align="center">
  Xiaocheng Fang, Zhengyao Ding, Guangkun Nie, Jieyi Cai, Yujie Xiao, Bo Liu<br/>
  Jiarui Jin, Haoyu Wang, Shun Huang, Ting Chen, Hongyan Li, Shenda Hong
</p>

</h5>
</div>

## Introduction: 
Cardiac Magnetic Resonance (CMR) imaging provides a comprehensive assessment of cardiac structure and function but remains constrained by high acquisition costs and reliance on expert annotations, limiting the availability of large-scale labeled datasets. In contrast, electrocardiograms (ECGs) are inexpensive, widely accessible, and offer a promising modality for conditioning the generative synthesis of cine CMR. To this end, we propose ECGFlowCMR, a novel ECG-to-CMR generative framework that integrates a Phase-Aware Masked Autoencoder (PA-MAE) and an Anatomy-Motion Disentangled Flow (AMDF) to address two fundamental challenges: (1) the cross-modal temporal mismatch between multi-beat ECG recordings and single-cycle CMR sequences, and (2) the anatomical observability gap due to the limited structural information inherent in ECGs. Extensive experiments on the UK Biobank and a proprietary clinical dataset demonstrate that ECGFlowCMR can generate realistic cine CMR sequences from ECG inputs, enabling scalable pretraining and improving performance on downstream cardiac disease classification and phenotype prediction tasks.

## 🫀 Cardiomyopathy Cases
We further showcase representative cardiomyopathy cases, including Dilated Cardiomyopathy (DCM) and Hypertrophic Cardiomyopathy (HCM), to evaluate generation fidelity in severe pathological states. Unlike conventional generative models that tend to regress toward a healthy mean anatomy, our method preserves disease-specific structural extremes. It successfully captures ventricular dilation, spherical remodeling, wall thinning, and global hypokinesis in DCM, while accurately reconstructing asymmetric septal hypertrophy and constrained ventricular volumes in HCM. These results highlight the ability of our model to preserve both anatomical morphology and cardiac dynamics under clinically challenging conditions.

## 🎬 Demo
<img src="videos/50.gif"/>
<img src="videos/388.gif"/>
<img src="videos/484.gif"/>
<img src="videos/1470.gif"/>
<img src="videos/3716.gif"/>
<img src="videos/4424.gif"/>

## ⚙️ ECGFlowCMR Training

1. Build environment from requirements.txt
2. Train the appropriate weights using your own dataset.

### Pretraining for Phase-Aware Masked Autoencoder:
```sh
python train_ecg.py
```
### Pretraining for 3D-VAE:
```sh
python train_disentangle_vae3d.py
```
### Training for Anatomy-Motion Disentangled Flow:
```sh
python train_flow.py
```
### Evaluation:
```sh
python evaluation.py
```

## 🚀 Quick Generation
You can use our pre-trained model weights for fast ECG to CMR.

### Our Model Checkpoints:
```sh
https://huggingface.co/XiaochengFang/ECGFlowCMR
```

### Generation:
```sh
python infer_flow.py
```

## 📝 Citation:
If you find our work interesting and helpful, please consider giving our repo a star. Additionally, if you would like to cite our work, please use the following format:
```sh
@article{fang2026ecgflowcmr,
  title={ECGFlowCMR: Pretraining with ECG-Generated Cine CMR Improves Cardiac Disease Classification and Phenotype Prediction},
  author={Fang, Xiaocheng and Ding, Zhengyao and Cai, Jieyi and Xiao, Yujie and Liu, Bo and Jin, Jiarui and Wang, Haoyu and Nie, Guangkun and Huang, Shun and Chen, Ting and others},
  journal={arXiv preprint arXiv:2601.20904},
  year={2026}
}
```

ECGFlowCMR

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README