CardioSpectrum: Comprehensive Myocardium Motion Analysis with 3D Deep Learning and Geometric Insights

Shahar Zuler1, Shai Tejman-Yarden2,3,4, Dan Raviv1,
1Tel Aviv University, Tel-Aviv, Israel
2Faculty of Medicine, Tel Aviv University, Ramat-Gan, Israel
3The Edmond J. Safra International Congenital Heart Center, Sheba Medical Center, Ramat-Gan, Israel
4The Engineering Medical Research Laboratory, Sheba Medical Center, Ramat-Gan, Israel
MICCAI 2024
Paper arXiv Code Poster
Data Paper Data Code
Teaser Image

CardioSpectrum effectively captures cardiac tangential movement (top), outperforming traditional methods that predominantly focus on local features, which often lead to inadequate results due to the aperture problem (bottom)

Abstract

The ability to map left ventricle (LV) myocardial motion using computed tomography angiography (CTA) is essential to diagnosing cardiovascular conditions and guiding interventional procedures. Due to their inherent locality, conventional neural networks typically have difficulty predicting subtle tangential movements, which considerably lessens the level of precision at which myocardium three-dimensional (3D) mapping can be performed. Using 3D optical flow techniques and Functional Maps (FMs), we present a comprehensive approach to address this problem. FMs are known for their capacity to capture global geometric features, thus providing a fuller understanding of 3D geometry. As an alternative to traditional segmentation-based priors, we employ surface-based two-dimensional (2D) constraints derived from spectral correspondence methods. Our 3D deep learning architecture, based on the ARFlow model, is optimized to handle complex 3D motion analysis tasks. By incorporating FMs, we can capture the subtle tangential movements of the myocardium surface precisely, hence significantly improving the accuracy of 3D mapping of the myocardium. The experimental results confirm the effectiveness of this method in enhancing myocardium motion analysis. This approach can contribute to improving cardiovascular diagnosis and treatment.

How does it work?

Mapping the movement of the LV myocardium accurately is vital in cardiac CT analysis. However, it's challenging due to the need to capture precise tangential movements. While traditional neural networks have made strides in analyzing cardiac movement, they often struggle with correctly predicting the tangential component, known as the aperture problem. To tackle this, we propose a holistic approach. We combine deep learning with spectral methods, specifically Functional Maps (FMs). FMs offer a global perspective, helping us understand myocardium motion comprehensively. CardioSpectrum method integrates 2D constraints derived from spectral correspondence methods, enhancing surface mapping accuracy.

The NN analyzes 3D image pairs of cardiac cycle timesteps such as systole and diastole, incorporating 2D constraints from ZoomOut. These constraints, derived from segmentations converted into meshes, result in a 3D optical flow.

The neural network analyzes 3D image pairs of cardiac cycle timesteps such as systole and diastole, incorporating 2D constraints from ZoomOut. These constraints, derived from segmentations converted into meshes, result in a 3D optical flow.

CardioSpectrum predicts more accurate deformation compared to the baselines as the torsion angle increases. Plot demonstrates mEPE (mean End-point Error) across the myocardium region.

TCardioSpectrum predicts more accurate deformation compared to the baselines as the torsion angle increases

Our use of real case data with synthetic deformations enables comprehensive comparison against deformation ground truth.

Our use of real case data with synthetic deformations enables comprehensive comparison against deformation ground truth

Acknowledgment

We would like to thank Netanel Nagar for his valuable contributions to the segmentation of data samples. His dedication, attention to detail, and commitment to excellence greatly enhanced the quality of this work.

BibTeX

If you find this work helpful, please cite our publication.

MICCAI 2024 proceedings:


    @InProceedings{ Zul_CardioSpectrum_MICCAI2024,
                 author = { Zuler, Shahar and Tejman-Yarden, Shai and Raviv, Dan },
                 title = { { CardioSpectrum: Comprehensive Myocardium Motion Analysis with 3D Deep Learning and Geometric Insights } }, 
                 booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2024},
                 year = {2024},
                 publisher = {Springer Nature Switzerland},
                 volume = { LNCS 15005 },
                 month = {October},
                 pages = { pending },
              }

arXiv preprint:


@misc{zuler2024cardiospectrumcomprehensivemyocardiummotion,
      title={CardioSpectrum: Comprehensive Myocardium Motion Analysis with 3D Deep Learning and Geometric Insights}, 
      author={Shahar Zuler and Shai Tejman-Yarden and Dan Raviv},
      year={2024},
      eprint={2407.03794},
      archivePrefix={arXiv},
      primaryClass={eess.IV},
      url={https://arxiv.org/abs/2407.03794}, 
}