Post-Infarction Risk Prediction with Mesh Classification Networks

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:10.1007/978-3-031-23443-9_27

Post-Infarction Risk Prediction with Mesh Classification Networks

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: Chapter in Book/Report/Conference proceeding › Conference paper › peer-review

1 Citation (Scopus)

Abstract

Post-myocardial infarction (MI) patients are at risk of major adverse cardiac events (MACE), with risk stratification primarily based on global image-based biomarkers, such as ejection fraction, in current clinical practice. However, these metrics neglect more subtle and localized shape differences in 3D cardiac anatomy and function, which limit predictive accuracy. In this work, we propose a novel geometric deep learning approach to directly predict MACE outcomes within 1 year after the infarction event from high-resolution 3D cardiac anatomy meshes. Its architecture is specifically designed for direct and efficient processing of surface mesh data with a hierarchical, multi-scale structure to enable both local and global feature learning. We evaluate the binary MACE prediction capabilities of the proposed mesh classification network on a multi-center dataset of post-MI patients. Our results show that the proposed method outperforms corresponding clinical benchmarks by ∼ 16% and ∼ 6% in terms of area under the receiver operating characteristic (AUROC) curve for 3D shape and 3D contraction inputs, respectively. Furthermore, we visually analyze both 3D cardiac shapes and 3D contraction patterns with regards to their MACE predictability and demonstrate how task-specific information learned by the network on a balanced dataset successfully generalizes to increasing levels of class imbalance. Finally, we compare our approach to both clinical and machine learning benchmarks on our original highly-imbalanced dataset of post-MI patients and find average improvements in AUROC scores of ∼ 9% and ∼ 3%, respectively.

Original language	English
Title of host publication	Statistical Atlases and Computational Models of the Heart. Regular and CMRxMotion Challenge Papers - 13th International Workshop, STACOM 2022, Held in Conjunction with MICCAI 2022, Revised Selected Papers
Editors	Oscar Camara, Esther Puyol-Antón, Avan Suinesiaputra, Alistair Young, Chen Qin, Maxime Sermesant, Shuo Wang
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	291-301
Number of pages	11
ISBN (Print)	9783031234422
DOIs	https://doi.org/10.1007/978-3-031-23443-9_27
Publication status	Published - 2022
Event	13th International Workshop on Statistical Atlases and Computational Models of the Heart, STACOM 2022, held in conjunction with the 25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022 - Singapore, Singapore Duration: 18 Sept 2022 → 18 Sept 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13593 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	13th International Workshop on Statistical Atlases and Computational Models of the Heart, STACOM 2022, held in conjunction with the 25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022
Country/Territory	Singapore
City	Singapore
Period	18/09/2022 → 18/09/2022

Keywords

3D cardiac contraction
3D cardiac shape
Cardiac MRI
Geometric deep learning
Graph neural networks
Major adverse cardiac event prediction
Mesh pooling
Post-myocardial infarction

Access to Document

10.1007/978-3-031-23443-9_27

Cite this

Beetz, M., Acero, J. C., Banerjee, A., Eitel, I., Zacur, E., Lange, T., Stiermaier, T., Evertz, R., Backhaus, S. J., Thiele, H., Bueno-Orovio, A., Lamata, P., Schuster, A., & Grau, V. (2022). Post-Infarction Risk Prediction with Mesh Classification Networks. In O. Camara, E. Puyol-Antón, A. Suinesiaputra, A. Young, C. Qin, M. Sermesant, & S. Wang (Eds.), Statistical Atlases and Computational Models of the Heart. Regular and CMRxMotion Challenge Papers - 13th International Workshop, STACOM 2022, Held in Conjunction with MICCAI 2022, Revised Selected Papers (pp. 291-301). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13593 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-23443-9_27

Beetz, Marcel ; Acero, Jorge Corral ; Banerjee, Abhirup et al. / Post-Infarction Risk Prediction with Mesh Classification Networks. Statistical Atlases and Computational Models of the Heart. Regular and CMRxMotion Challenge Papers - 13th International Workshop, STACOM 2022, Held in Conjunction with MICCAI 2022, Revised Selected Papers. editor / Oscar Camara ; Esther Puyol-Antón ; Avan Suinesiaputra ; Alistair Young ; Chen Qin ; Maxime Sermesant ; Shuo Wang. Springer Science and Business Media Deutschland GmbH, 2022. pp. 291-301 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inbook{67de78a32d32480cb7e78b0b9abc3ce7,

title = "Post-Infarction Risk Prediction with Mesh Classification Networks",

abstract = "Post-myocardial infarction (MI) patients are at risk of major adverse cardiac events (MACE), with risk stratification primarily based on global image-based biomarkers, such as ejection fraction, in current clinical practice. However, these metrics neglect more subtle and localized shape differences in 3D cardiac anatomy and function, which limit predictive accuracy. In this work, we propose a novel geometric deep learning approach to directly predict MACE outcomes within 1 year after the infarction event from high-resolution 3D cardiac anatomy meshes. Its architecture is specifically designed for direct and efficient processing of surface mesh data with a hierarchical, multi-scale structure to enable both local and global feature learning. We evaluate the binary MACE prediction capabilities of the proposed mesh classification network on a multi-center dataset of post-MI patients. Our results show that the proposed method outperforms corresponding clinical benchmarks by ∼ 16% and ∼ 6% in terms of area under the receiver operating characteristic (AUROC) curve for 3D shape and 3D contraction inputs, respectively. Furthermore, we visually analyze both 3D cardiac shapes and 3D contraction patterns with regards to their MACE predictability and demonstrate how task-specific information learned by the network on a balanced dataset successfully generalizes to increasing levels of class imbalance. Finally, we compare our approach to both clinical and machine learning benchmarks on our original highly-imbalanced dataset of post-MI patients and find average improvements in AUROC scores of ∼ 9% and ∼ 3%, respectively.",

keywords = "3D cardiac contraction, 3D cardiac shape, Cardiac MRI, Geometric deep learning, Graph neural networks, Major adverse cardiac event prediction, Mesh pooling, Post-myocardial infarction",

author = "Marcel Beetz and Acero, {Jorge Corral} 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: Acknowledgments. The authors express no conflict of interest. The work of MB is supported by the Stiftung der Deutschen Wirtschaft (Foundation of German Business). AB is a Royal Society University Research Fellow and is supported by the Royal Society (Grant No. URF\R1\221314). The work of AB and VG is supported by the British Heart Foundation (BHF) Project under Grant PG/20/21/35082. The work of VG is supported by the CompBioMed 2 Centre of Excellence in Computational Biomedicine (European Commission Horizon 2020 research and innovation programme, grant agreement No. 823712). The work of JCA is supported by the EU{\textquoteright}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. ABO holds a BHF Intermediate Basic Science Research Fellowship (FS/17/22/32644). The work is 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: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 13th International Workshop on Statistical Atlases and Computational Models of the Heart, STACOM 2022, held in conjunction with the 25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022 ; Conference date: 18-09-2022 Through 18-09-2022",

year = "2022",

doi = "10.1007/978-3-031-23443-9_27",

language = "English",

isbn = "9783031234422",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "291--301",

editor = "Oscar Camara and Esther Puyol-Ant{\'o}n and Avan Suinesiaputra and Alistair Young and Chen Qin and Maxime Sermesant and Shuo Wang",

booktitle = "Statistical Atlases and Computational Models of the Heart. Regular and CMRxMotion Challenge Papers - 13th International Workshop, STACOM 2022, Held in Conjunction with MICCAI 2022, Revised Selected Papers",

address = "Germany",

}

Beetz, M, Acero, JC, Banerjee, A, Eitel, I, Zacur, E, Lange, T, Stiermaier, T, Evertz, R, Backhaus, SJ, Thiele, H, Bueno-Orovio, A, Lamata, P, Schuster, A & Grau, V 2022, Post-Infarction Risk Prediction with Mesh Classification Networks. in O Camara, E Puyol-Antón, A Suinesiaputra, A Young, C Qin, M Sermesant & S Wang (eds), Statistical Atlases and Computational Models of the Heart. Regular and CMRxMotion Challenge Papers - 13th International Workshop, STACOM 2022, Held in Conjunction with MICCAI 2022, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13593 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 291-301, 13th International Workshop on Statistical Atlases and Computational Models of the Heart, STACOM 2022, held in conjunction with the 25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022, Singapore, Singapore, 18/09/2022. https://doi.org/10.1007/978-3-031-23443-9_27

Post-Infarction Risk Prediction with Mesh Classification Networks. / Beetz, Marcel; Acero, Jorge Corral; Banerjee, Abhirup et al.
Statistical Atlases and Computational Models of the Heart. Regular and CMRxMotion Challenge Papers - 13th International Workshop, STACOM 2022, Held in Conjunction with MICCAI 2022, Revised Selected Papers. ed. / Oscar Camara; Esther Puyol-Antón; Avan Suinesiaputra; Alistair Young; Chen Qin; Maxime Sermesant; Shuo Wang. Springer Science and Business Media Deutschland GmbH, 2022. p. 291-301 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13593 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › peer-review

TY - CHAP

T1 - Post-Infarction Risk Prediction with Mesh Classification Networks

AU - Beetz, Marcel

AU - Acero, Jorge Corral

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: Acknowledgments. The authors express no conflict of interest. The work of MB is supported by the Stiftung der Deutschen Wirtschaft (Foundation of German Business). AB is a Royal Society University Research Fellow and is supported by the Royal Society (Grant No. URF\R1\221314). The work of AB and VG is supported by the British Heart Foundation (BHF) Project under Grant PG/20/21/35082. The work of VG is supported by the CompBioMed 2 Centre of Excellence in Computational Biomedicine (European Commission Horizon 2020 research and innovation programme, grant agreement No. 823712). The work of JCA is 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. ABO holds a BHF Intermediate Basic Science Research Fellowship (FS/17/22/32644). The work is 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: © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

PY - 2022

Y1 - 2022

N2 - Post-myocardial infarction (MI) patients are at risk of major adverse cardiac events (MACE), with risk stratification primarily based on global image-based biomarkers, such as ejection fraction, in current clinical practice. However, these metrics neglect more subtle and localized shape differences in 3D cardiac anatomy and function, which limit predictive accuracy. In this work, we propose a novel geometric deep learning approach to directly predict MACE outcomes within 1 year after the infarction event from high-resolution 3D cardiac anatomy meshes. Its architecture is specifically designed for direct and efficient processing of surface mesh data with a hierarchical, multi-scale structure to enable both local and global feature learning. We evaluate the binary MACE prediction capabilities of the proposed mesh classification network on a multi-center dataset of post-MI patients. Our results show that the proposed method outperforms corresponding clinical benchmarks by ∼ 16% and ∼ 6% in terms of area under the receiver operating characteristic (AUROC) curve for 3D shape and 3D contraction inputs, respectively. Furthermore, we visually analyze both 3D cardiac shapes and 3D contraction patterns with regards to their MACE predictability and demonstrate how task-specific information learned by the network on a balanced dataset successfully generalizes to increasing levels of class imbalance. Finally, we compare our approach to both clinical and machine learning benchmarks on our original highly-imbalanced dataset of post-MI patients and find average improvements in AUROC scores of ∼ 9% and ∼ 3%, respectively.

AB - Post-myocardial infarction (MI) patients are at risk of major adverse cardiac events (MACE), with risk stratification primarily based on global image-based biomarkers, such as ejection fraction, in current clinical practice. However, these metrics neglect more subtle and localized shape differences in 3D cardiac anatomy and function, which limit predictive accuracy. In this work, we propose a novel geometric deep learning approach to directly predict MACE outcomes within 1 year after the infarction event from high-resolution 3D cardiac anatomy meshes. Its architecture is specifically designed for direct and efficient processing of surface mesh data with a hierarchical, multi-scale structure to enable both local and global feature learning. We evaluate the binary MACE prediction capabilities of the proposed mesh classification network on a multi-center dataset of post-MI patients. Our results show that the proposed method outperforms corresponding clinical benchmarks by ∼ 16% and ∼ 6% in terms of area under the receiver operating characteristic (AUROC) curve for 3D shape and 3D contraction inputs, respectively. Furthermore, we visually analyze both 3D cardiac shapes and 3D contraction patterns with regards to their MACE predictability and demonstrate how task-specific information learned by the network on a balanced dataset successfully generalizes to increasing levels of class imbalance. Finally, we compare our approach to both clinical and machine learning benchmarks on our original highly-imbalanced dataset of post-MI patients and find average improvements in AUROC scores of ∼ 9% and ∼ 3%, respectively.

KW - 3D cardiac contraction

KW - 3D cardiac shape

KW - Cardiac MRI

KW - Geometric deep learning

KW - Graph neural networks

KW - Major adverse cardiac event prediction

KW - Mesh pooling

KW - Post-myocardial infarction

UR - http://www.scopus.com/inward/record.url?scp=85148034771&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-23443-9_27

DO - 10.1007/978-3-031-23443-9_27

M3 - Conference paper

AN - SCOPUS:85148034771

SN - 9783031234422

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 291

EP - 301

BT - Statistical Atlases and Computational Models of the Heart. Regular and CMRxMotion Challenge Papers - 13th International Workshop, STACOM 2022, Held in Conjunction with MICCAI 2022, Revised Selected Papers

A2 - Camara, Oscar

A2 - Puyol-Antón, Esther

A2 - Suinesiaputra, Avan

A2 - Young, Alistair

A2 - Qin, Chen

A2 - Sermesant, Maxime

A2 - Wang, Shuo

PB - Springer Science and Business Media Deutschland GmbH

T2 - 13th International Workshop on Statistical Atlases and Computational Models of the Heart, STACOM 2022, held in conjunction with the 25th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2022

Y2 - 18 September 2022 through 18 September 2022

ER -

Beetz M, Acero JC, Banerjee A, Eitel I, Zacur E, Lange T et al. Post-Infarction Risk Prediction with Mesh Classification Networks. In Camara O, Puyol-Antón E, Suinesiaputra A, Young A, Qin C, Sermesant M, Wang S, editors, Statistical Atlases and Computational Models of the Heart. Regular and CMRxMotion Challenge Papers - 13th International Workshop, STACOM 2022, Held in Conjunction with MICCAI 2022, Revised Selected Papers. Springer Science and Business Media Deutschland GmbH. 2022. p. 291-301. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-23443-9_27