A deep learning approach for fully automated cardiac shape modeling in tetralogy of Fallot

Sachin Govil; Brendan T Crabb; Yu Deng; Laura Dal Toso; Esther Puyol-Antón; Kuberan Pushparajah; Sanjeet Hegde; James C Perry; Jeffrey H Omens; Albert Hsiao; Alistair A Young; Andrew D McCulloch

doi:10.1186/s12968-023-00924-1

A deep learning approach for fully automated cardiac shape modeling in tetralogy of Fallot

Sachin Govil, Brendan T Crabb, Yu Deng, Laura Dal Toso, Esther Puyol-Antón, Kuberan Pushparajah, Sanjeet Hegde, James C Perry, Jeffrey H Omens, Albert Hsiao, Alistair A Young, Andrew D McCulloch

King's College Hospital

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

9 Citations (Scopus)

Abstract

BACKGROUND: Cardiac shape modeling is a useful computational tool that has provided quantitative insights into the mechanisms underlying dysfunction in heart disease. The manual input and time required to make cardiac shape models, however, limits their clinical utility. Here we present an end-to-end pipeline that uses deep learning for automated view classification, slice selection, phase selection, anatomical landmark localization, and myocardial image segmentation for the automated generation of three-dimensional, biventricular shape models. With this approach, we aim to make cardiac shape modeling a more robust and broadly applicable tool that has processing times consistent with clinical workflows.

METHODS: Cardiovascular magnetic resonance (CMR) images from a cohort of 123 patients with repaired tetralogy of Fallot (rTOF) from two internal sites were used to train and validate each step in the automated pipeline. The complete automated pipeline was tested using CMR images from a cohort of 12 rTOF patients from an internal site and 18 rTOF patients from an external site. Manually and automatically generated shape models from the test set were compared using Euclidean projection distances, global ventricular measurements, and atlas-based shape mode scores.

RESULTS: The mean absolute error (MAE) between manually and automatically generated shape models in the test set was similar to the voxel resolution of the original CMR images for end-diastolic models (MAE = 1.9 ± 0.5 mm) and end-systolic models (MAE = 2.1 ± 0.7 mm). Global ventricular measurements computed from automated models were in good agreement with those computed from manual models. The average mean absolute difference in shape mode Z-score between manually and automatically generated models was 0.5 standard deviations for the first 20 modes of a reference statistical shape atlas.

CONCLUSIONS: Using deep learning, accurate three-dimensional, biventricular shape models can be reliably created. This fully automated end-to-end approach dramatically reduces the manual input required to create shape models, thereby enabling the rapid analysis of large-scale datasets and the potential to deploy statistical atlas-based analyses in point-of-care clinical settings. Training data and networks are available from cardiacatlas.org.

Original language	English
Article number	15
Journal	Journal of Cardiovascular Magnetic Resonance
Volume	25
Issue number	1
DOIs	https://doi.org/10.1186/s12968-023-00924-1
Publication status	Published - 27 Feb 2023

Keywords

Cardiovascular magnetic resonance (CMR)
Congenital heart disease
Deep learning
Image segmentation
Shape modeling

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1186/s12968-023-00924-1

Cite this

Govil, S., Crabb, B. T., Deng, Y., Dal Toso, L., Puyol-Antón, E., Pushparajah, K., Hegde, S., Perry, J. C., Omens, J. H., Hsiao, A., Young, A. A., & McCulloch, A. D. (2023). A deep learning approach for fully automated cardiac shape modeling in tetralogy of Fallot. Journal of Cardiovascular Magnetic Resonance, 25(1), Article 15. https://doi.org/10.1186/s12968-023-00924-1

@article{41fa201efde34db19b1becc2984e53ad,

title = "A deep learning approach for fully automated cardiac shape modeling in tetralogy of Fallot",

abstract = "BACKGROUND: Cardiac shape modeling is a useful computational tool that has provided quantitative insights into the mechanisms underlying dysfunction in heart disease. The manual input and time required to make cardiac shape models, however, limits their clinical utility. Here we present an end-to-end pipeline that uses deep learning for automated view classification, slice selection, phase selection, anatomical landmark localization, and myocardial image segmentation for the automated generation of three-dimensional, biventricular shape models. With this approach, we aim to make cardiac shape modeling a more robust and broadly applicable tool that has processing times consistent with clinical workflows.METHODS: Cardiovascular magnetic resonance (CMR) images from a cohort of 123 patients with repaired tetralogy of Fallot (rTOF) from two internal sites were used to train and validate each step in the automated pipeline. The complete automated pipeline was tested using CMR images from a cohort of 12 rTOF patients from an internal site and 18 rTOF patients from an external site. Manually and automatically generated shape models from the test set were compared using Euclidean projection distances, global ventricular measurements, and atlas-based shape mode scores.RESULTS: The mean absolute error (MAE) between manually and automatically generated shape models in the test set was similar to the voxel resolution of the original CMR images for end-diastolic models (MAE = 1.9 ± 0.5 mm) and end-systolic models (MAE = 2.1 ± 0.7 mm). Global ventricular measurements computed from automated models were in good agreement with those computed from manual models. The average mean absolute difference in shape mode Z-score between manually and automatically generated models was 0.5 standard deviations for the first 20 modes of a reference statistical shape atlas.CONCLUSIONS: Using deep learning, accurate three-dimensional, biventricular shape models can be reliably created. This fully automated end-to-end approach dramatically reduces the manual input required to create shape models, thereby enabling the rapid analysis of large-scale datasets and the potential to deploy statistical atlas-based analyses in point-of-care clinical settings. Training data and networks are available from cardiacatlas.org.",

keywords = "Cardiovascular magnetic resonance (CMR), Congenital heart disease, Deep learning, Image segmentation, Shape modeling",

author = "Sachin Govil and Crabb, {Brendan T} and Yu Deng and {Dal Toso}, Laura and Esther Puyol-Ant{\'o}n and Kuberan Pushparajah and Sanjeet Hegde and Perry, {James C} and Omens, {Jeffrey H} and Albert Hsiao and Young, {Alistair A} and McCulloch, {Andrew D}",

note = "Funding Information: Funding was provided by National Institutes of Health R01HL121754, American Heart Association 19AIML35120034, and the Saving Tiny Hearts Society. SG acknowledges National Institutes of Health NHLBI T32HL105373. AY and LDT acknowledge Health Research Council of New Zealand 17/234 and Wellcome ESPCR Centre for Medical Engineering at King{\textquoteright}s College London WT203148/Z/16/Z. Funding Information: We would like to acknowledge Fernando Ramirez, a research associate at Rady Children{\textquoteright}s Hospital San Diego, for involvement in data collection and data transfer. Publisher Copyright: {\textcopyright} 2023, The Author(s).",

year = "2023",

month = feb,

day = "27",

doi = "10.1186/s12968-023-00924-1",

language = "English",

volume = "25",

journal = "Journal of Cardiovascular Magnetic Resonance",

issn = "1097-6647",

publisher = "BioMed Central",

number = "1",

}

TY - JOUR

T1 - A deep learning approach for fully automated cardiac shape modeling in tetralogy of Fallot

AU - Govil, Sachin

AU - Crabb, Brendan T

AU - Deng, Yu

AU - Dal Toso, Laura

AU - Puyol-Antón, Esther

AU - Pushparajah, Kuberan

AU - Hegde, Sanjeet

AU - Perry, James C

AU - Omens, Jeffrey H

AU - Hsiao, Albert

AU - Young, Alistair A

AU - McCulloch, Andrew D

N1 - Funding Information: Funding was provided by National Institutes of Health R01HL121754, American Heart Association 19AIML35120034, and the Saving Tiny Hearts Society. SG acknowledges National Institutes of Health NHLBI T32HL105373. AY and LDT acknowledge Health Research Council of New Zealand 17/234 and Wellcome ESPCR Centre for Medical Engineering at King’s College London WT203148/Z/16/Z. Funding Information: We would like to acknowledge Fernando Ramirez, a research associate at Rady Children’s Hospital San Diego, for involvement in data collection and data transfer. Publisher Copyright: © 2023, The Author(s).

PY - 2023/2/27

Y1 - 2023/2/27

N2 - BACKGROUND: Cardiac shape modeling is a useful computational tool that has provided quantitative insights into the mechanisms underlying dysfunction in heart disease. The manual input and time required to make cardiac shape models, however, limits their clinical utility. Here we present an end-to-end pipeline that uses deep learning for automated view classification, slice selection, phase selection, anatomical landmark localization, and myocardial image segmentation for the automated generation of three-dimensional, biventricular shape models. With this approach, we aim to make cardiac shape modeling a more robust and broadly applicable tool that has processing times consistent with clinical workflows.METHODS: Cardiovascular magnetic resonance (CMR) images from a cohort of 123 patients with repaired tetralogy of Fallot (rTOF) from two internal sites were used to train and validate each step in the automated pipeline. The complete automated pipeline was tested using CMR images from a cohort of 12 rTOF patients from an internal site and 18 rTOF patients from an external site. Manually and automatically generated shape models from the test set were compared using Euclidean projection distances, global ventricular measurements, and atlas-based shape mode scores.RESULTS: The mean absolute error (MAE) between manually and automatically generated shape models in the test set was similar to the voxel resolution of the original CMR images for end-diastolic models (MAE = 1.9 ± 0.5 mm) and end-systolic models (MAE = 2.1 ± 0.7 mm). Global ventricular measurements computed from automated models were in good agreement with those computed from manual models. The average mean absolute difference in shape mode Z-score between manually and automatically generated models was 0.5 standard deviations for the first 20 modes of a reference statistical shape atlas.CONCLUSIONS: Using deep learning, accurate three-dimensional, biventricular shape models can be reliably created. This fully automated end-to-end approach dramatically reduces the manual input required to create shape models, thereby enabling the rapid analysis of large-scale datasets and the potential to deploy statistical atlas-based analyses in point-of-care clinical settings. Training data and networks are available from cardiacatlas.org.

AB - BACKGROUND: Cardiac shape modeling is a useful computational tool that has provided quantitative insights into the mechanisms underlying dysfunction in heart disease. The manual input and time required to make cardiac shape models, however, limits their clinical utility. Here we present an end-to-end pipeline that uses deep learning for automated view classification, slice selection, phase selection, anatomical landmark localization, and myocardial image segmentation for the automated generation of three-dimensional, biventricular shape models. With this approach, we aim to make cardiac shape modeling a more robust and broadly applicable tool that has processing times consistent with clinical workflows.METHODS: Cardiovascular magnetic resonance (CMR) images from a cohort of 123 patients with repaired tetralogy of Fallot (rTOF) from two internal sites were used to train and validate each step in the automated pipeline. The complete automated pipeline was tested using CMR images from a cohort of 12 rTOF patients from an internal site and 18 rTOF patients from an external site. Manually and automatically generated shape models from the test set were compared using Euclidean projection distances, global ventricular measurements, and atlas-based shape mode scores.RESULTS: The mean absolute error (MAE) between manually and automatically generated shape models in the test set was similar to the voxel resolution of the original CMR images for end-diastolic models (MAE = 1.9 ± 0.5 mm) and end-systolic models (MAE = 2.1 ± 0.7 mm). Global ventricular measurements computed from automated models were in good agreement with those computed from manual models. The average mean absolute difference in shape mode Z-score between manually and automatically generated models was 0.5 standard deviations for the first 20 modes of a reference statistical shape atlas.CONCLUSIONS: Using deep learning, accurate three-dimensional, biventricular shape models can be reliably created. This fully automated end-to-end approach dramatically reduces the manual input required to create shape models, thereby enabling the rapid analysis of large-scale datasets and the potential to deploy statistical atlas-based analyses in point-of-care clinical settings. Training data and networks are available from cardiacatlas.org.

KW - Cardiovascular magnetic resonance (CMR)

KW - Congenital heart disease

KW - Deep learning

KW - Image segmentation

KW - Shape modeling

UR - http://www.scopus.com/inward/record.url?scp=85148970309&partnerID=8YFLogxK

U2 - 10.1186/s12968-023-00924-1

DO - 10.1186/s12968-023-00924-1

M3 - Article

C2 - 36849960

AN - SCOPUS:85148970309

SN - 1097-6647

VL - 25

JO - Journal of Cardiovascular Magnetic Resonance

JF - Journal of Cardiovascular Magnetic Resonance

IS - 1

M1 - 15

ER -

A deep learning approach for fully automated cardiac shape modeling in tetralogy of Fallot

Abstract

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this