# Voxel2Hemodynamics: An End-to-end Deep Learning Method for Predicting Coronary Artery Hemodynamics

arXiv: https://arxiv.org/abs/2305.19107 · HTML: https://mlsanity.com/paper/2305.19107/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 4 of its 24 references that are in the map.
- First posted on arXiv 30 May 2023.

## Builds on (most cited first, 4 of 4)

- [Semi-Supervised Classification with Graph Convolutional Networks](https://mlsanity.com/paper/1609.02907/) (2016)
- [PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space](https://mlsanity.com/paper/1706.02413/) (2017)
- [3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation](https://mlsanity.com/paper/1606.06650/) (2016)
- [Pixel2Mesh: Generating 3D Mesh Models from Single RGB Images](https://mlsanity.com/paper/1804.01654/) (2018)

## Built on by (most cited first, 2 of 2)

- [Graph Neural Network for Cerebral Blood Flow Prediction With Clinical Datasets](https://mlsanity.com/paper/2411.17971/) (2024)
- [Stable Long-Horizon Spatiotemporal Prediction on Meshes Using Latent Multiscale Recurrent Graph Neural Networks](https://mlsanity.com/paper/2602.18146/) (2026)

## Abstract (arXiv)

Local hemodynamic forces play an important role in determining the functional significance of coronary arterial stenosis and understanding the mechanism of coronary disease progression. Computational fluid dynamics (CFD) have been widely performed to simulate hemodynamics non-invasively from coronary computed tomography angiography (CCTA) images. However, accurate computational analysis is still limited by the complex construction of patient-specific modeling and time-consuming computation. In this work, we proposed an end-to-end deep learning framework, which could predict the coronary artery hemodynamics from CCTA images. The model was trained on the hemodynamic data obtained from 3D simulations of synthetic and real datasets. Extensive experiments demonstrated that the predicted hemdynamic distributions by our method agreed well with the CFD-derived results. Quantitatively, the proposed method has the capability of predicting the fractional flow reserve with an average error of 0.5% and 2.5% for the synthetic dataset and real dataset, respectively. Particularly, our method achieved much better accuracy for the real dataset compared to PointNet++ with the point cloud input. This study demonstrates the feasibility and great potential of our end-to-end deep learning method as a fast and accurate approach for hemodynamic analysis.