Interpretable cardiac anatomy modeling using variational mesh autoencoders

Marcel Beetz; Jorge Corral Acero; Abhirup Banerjee; Ingo Eitel; Ernesto Zacur; Torben Lange; Thomas Stiermaier; Ruben Evertz; Sören J. Backhaus; Holger Thiele; Alfonso Bueno-Orovio; Pablo Lamata; Andreas Schuster; Vicente Grau

doi:10.3389/fcvm.2022.983868

Interpretable cardiac anatomy modeling using variational mesh autoencoders

Marcel Beetz^*, Jorge Corral Acero, Abhirup Banerjee, Ingo Eitel, Ernesto Zacur, Torben Lange, Thomas Stiermaier, Ruben Evertz, Sören J. Backhaus, Holger Thiele, Alfonso Bueno-Orovio, Pablo Lamata, Andreas Schuster, Vicente Grau

^*Corresponding author for this work

Biomedical Engineering Department

Research output: Contribution to journal › Article › peer-review

14 Citations (Scopus)

Abstract

Cardiac anatomy and function vary considerably across the human population with important implications for clinical diagnosis and treatment planning. Consequently, many computer-based approaches have been developed to capture this variability for a wide range of applications, including explainable cardiac disease detection and prediction, dimensionality reduction, cardiac shape analysis, and the generation of virtual heart populations. In this work, we propose a variational mesh autoencoder (mesh VAE) as a novel geometric deep learning approach to model such population-wide variations in cardiac shapes. It embeds multi-scale graph convolutions and mesh pooling layers in a hierarchical VAE framework to enable direct processing of surface mesh representations of the cardiac anatomy in an efficient manner. The proposed mesh VAE achieves low reconstruction errors on a dataset of 3D cardiac meshes from over 1,000 patients with acute myocardial infarction, with mean surface distances between input and reconstructed meshes below the underlying image resolution. We also find that it outperforms a voxelgrid-based deep learning benchmark in terms of both mean surface distance and Hausdorff distance while requiring considerably less memory. Furthermore, we explore the quality and interpretability of the mesh VAE's latent space and showcase its ability to improve the prediction of major adverse cardiac events over a clinical benchmark. Finally, we investigate the method's ability to generate realistic virtual populations of cardiac anatomies and find good alignment between the synthesized and gold standard mesh populations in terms of multiple clinical metrics.

Original language	English
Article number	983868
Journal	Frontiers in Cardiovascular Medicine
Volume	9
DOIs	https://doi.org/10.3389/fcvm.2022.983868
Publication status	Published - 22 Dec 2022

Keywords

3D ventricular shape analysis
acute myocardial infarction
clinical outcome prediction
geometric deep learning
graph neural networks
major adverse cardiac events
mesh VAE
virtual anatomy generation

Access to Document

10.3389/fcvm.2022.983868

Cite this

@article{44414e4b7b64493580b5ee3b0df84643,

title = "Interpretable cardiac anatomy modeling using variational mesh autoencoders",

abstract = "Cardiac anatomy and function vary considerably across the human population with important implications for clinical diagnosis and treatment planning. Consequently, many computer-based approaches have been developed to capture this variability for a wide range of applications, including explainable cardiac disease detection and prediction, dimensionality reduction, cardiac shape analysis, and the generation of virtual heart populations. In this work, we propose a variational mesh autoencoder (mesh VAE) as a novel geometric deep learning approach to model such population-wide variations in cardiac shapes. It embeds multi-scale graph convolutions and mesh pooling layers in a hierarchical VAE framework to enable direct processing of surface mesh representations of the cardiac anatomy in an efficient manner. The proposed mesh VAE achieves low reconstruction errors on a dataset of 3D cardiac meshes from over 1,000 patients with acute myocardial infarction, with mean surface distances between input and reconstructed meshes below the underlying image resolution. We also find that it outperforms a voxelgrid-based deep learning benchmark in terms of both mean surface distance and Hausdorff distance while requiring considerably less memory. Furthermore, we explore the quality and interpretability of the mesh VAE's latent space and showcase its ability to improve the prediction of major adverse cardiac events over a clinical benchmark. Finally, we investigate the method's ability to generate realistic virtual populations of cardiac anatomies and find good alignment between the synthesized and gold standard mesh populations in terms of multiple clinical metrics.",

keywords = "3D ventricular shape analysis, acute myocardial infarction, clinical outcome prediction, geometric deep learning, graph neural networks, major adverse cardiac events, mesh VAE, virtual anatomy generation",

author = "Marcel Beetz and {Corral Acero}, Jorge and Abhirup Banerjee and Ingo Eitel and Ernesto Zacur and Torben Lange and Thomas Stiermaier and Ruben Evertz and Backhaus, {S{\"o}ren J.} and Holger Thiele and Alfonso Bueno-Orovio and Pablo Lamata and Andreas Schuster and Vicente Grau",

note = "Funding Information: The work of MB was supported by the Stiftung der Deutschen Wirtschaft (Foundation of German Business). The work of JC was supported by the EU's Horizon 2020 research and innovation program under the Marie Sklodowska-Curie (g.a. 764738) and the EPSRC Impact Acceleration Account (D4D00010 DF48.01), funded by UK Research and Innovation. AB is a Royal Society University Research Fellow and supported by the Royal Society (Grant No. URF\R1\221314). The work of AB and VG was supported by the British Heart Foundation (BHF) Project under Grant HSR01230. The work of VG was supported by the CompBioMed 2 Centre of Excellence in Computational Biomedicine (European Commission Horizon 2020 research and innovation programme, grant agreement No. 823712). AB-O holds a BHF Intermediate Basic Science Research Fellowship (FS/17/22/32644). The work was also supported by the German Center for Cardiovascular Research, the British Heart Foundation (PG/16/75/32383), and the Wellcome Trust (209450/Z/17). Publisher Copyright: Copyright {\textcopyright} 2022 Beetz, Corral Acero, Banerjee, Eitel, Zacur, Lange, Stiermaier, Evertz, Backhaus, Thiele, Bueno-Orovio, Lamata, Schuster and Grau.",

year = "2022",

month = dec,

day = "22",

doi = "10.3389/fcvm.2022.983868",

language = "English",

volume = "9",

journal = "Frontiers in Cardiovascular Medicine",

issn = "2297-055X",

publisher = "Frontiers Media S.A.",

}

TY - JOUR

T1 - Interpretable cardiac anatomy modeling using variational mesh autoencoders

AU - Beetz, Marcel

AU - Corral Acero, Jorge

AU - Banerjee, Abhirup

AU - Eitel, Ingo

AU - Zacur, Ernesto

AU - Lange, Torben

AU - Stiermaier, Thomas

AU - Evertz, Ruben

AU - Backhaus, Sören J.

AU - Thiele, Holger

AU - Bueno-Orovio, Alfonso

AU - Lamata, Pablo

AU - Schuster, Andreas

AU - Grau, Vicente

N1 - Funding Information: The work of MB was supported by the Stiftung der Deutschen Wirtschaft (Foundation of German Business). The work of JC was supported by the EU's Horizon 2020 research and innovation program under the Marie Sklodowska-Curie (g.a. 764738) and the EPSRC Impact Acceleration Account (D4D00010 DF48.01), funded by UK Research and Innovation. AB is a Royal Society University Research Fellow and supported by the Royal Society (Grant No. URF\R1\221314). The work of AB and VG was supported by the British Heart Foundation (BHF) Project under Grant HSR01230. The work of VG was supported by the CompBioMed 2 Centre of Excellence in Computational Biomedicine (European Commission Horizon 2020 research and innovation programme, grant agreement No. 823712). AB-O holds a BHF Intermediate Basic Science Research Fellowship (FS/17/22/32644). The work was also supported by the German Center for Cardiovascular Research, the British Heart Foundation (PG/16/75/32383), and the Wellcome Trust (209450/Z/17). Publisher Copyright: Copyright © 2022 Beetz, Corral Acero, Banerjee, Eitel, Zacur, Lange, Stiermaier, Evertz, Backhaus, Thiele, Bueno-Orovio, Lamata, Schuster and Grau.

PY - 2022/12/22

Y1 - 2022/12/22

N2 - Cardiac anatomy and function vary considerably across the human population with important implications for clinical diagnosis and treatment planning. Consequently, many computer-based approaches have been developed to capture this variability for a wide range of applications, including explainable cardiac disease detection and prediction, dimensionality reduction, cardiac shape analysis, and the generation of virtual heart populations. In this work, we propose a variational mesh autoencoder (mesh VAE) as a novel geometric deep learning approach to model such population-wide variations in cardiac shapes. It embeds multi-scale graph convolutions and mesh pooling layers in a hierarchical VAE framework to enable direct processing of surface mesh representations of the cardiac anatomy in an efficient manner. The proposed mesh VAE achieves low reconstruction errors on a dataset of 3D cardiac meshes from over 1,000 patients with acute myocardial infarction, with mean surface distances between input and reconstructed meshes below the underlying image resolution. We also find that it outperforms a voxelgrid-based deep learning benchmark in terms of both mean surface distance and Hausdorff distance while requiring considerably less memory. Furthermore, we explore the quality and interpretability of the mesh VAE's latent space and showcase its ability to improve the prediction of major adverse cardiac events over a clinical benchmark. Finally, we investigate the method's ability to generate realistic virtual populations of cardiac anatomies and find good alignment between the synthesized and gold standard mesh populations in terms of multiple clinical metrics.

AB - Cardiac anatomy and function vary considerably across the human population with important implications for clinical diagnosis and treatment planning. Consequently, many computer-based approaches have been developed to capture this variability for a wide range of applications, including explainable cardiac disease detection and prediction, dimensionality reduction, cardiac shape analysis, and the generation of virtual heart populations. In this work, we propose a variational mesh autoencoder (mesh VAE) as a novel geometric deep learning approach to model such population-wide variations in cardiac shapes. It embeds multi-scale graph convolutions and mesh pooling layers in a hierarchical VAE framework to enable direct processing of surface mesh representations of the cardiac anatomy in an efficient manner. The proposed mesh VAE achieves low reconstruction errors on a dataset of 3D cardiac meshes from over 1,000 patients with acute myocardial infarction, with mean surface distances between input and reconstructed meshes below the underlying image resolution. We also find that it outperforms a voxelgrid-based deep learning benchmark in terms of both mean surface distance and Hausdorff distance while requiring considerably less memory. Furthermore, we explore the quality and interpretability of the mesh VAE's latent space and showcase its ability to improve the prediction of major adverse cardiac events over a clinical benchmark. Finally, we investigate the method's ability to generate realistic virtual populations of cardiac anatomies and find good alignment between the synthesized and gold standard mesh populations in terms of multiple clinical metrics.

KW - 3D ventricular shape analysis

KW - acute myocardial infarction

KW - clinical outcome prediction

KW - geometric deep learning

KW - graph neural networks

KW - major adverse cardiac events

KW - mesh VAE

KW - virtual anatomy generation

UR - http://www.scopus.com/inward/record.url?scp=85145739837&partnerID=8YFLogxK

U2 - 10.3389/fcvm.2022.983868

DO - 10.3389/fcvm.2022.983868

M3 - Article

AN - SCOPUS:85145739837

SN - 2297-055X

VL - 9

JO - Frontiers in Cardiovascular Medicine

JF - Frontiers in Cardiovascular Medicine

M1 - 983868

ER -

Interpretable cardiac anatomy modeling using variational mesh autoencoders

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this