Impact of training data composition on the generalizability of convolutional neural network aortic cross-section segmentation in four-dimensional magnetic resonance flow imaging

ABSTRACT: Background: Four-dimensional cardiovascular magnetic resonance flow imaging (4D flow CMR) plays an important role in assessing cardiovascular diseases. However, the manual or semi-automatic segmentation of aortic vessel boundaries in 4D flow data introduces variability and limits the repr...

Full description

Saved in:
Bibliographic Details
Main Authors: Chiara Manini, Markus Hüllebrand, Lars Walczak, Sarah Nordmeyer, Lina Jarmatz, Titus Kuehne, Heiko Stern, Christian Meierhofer, Andreas Harloff, Jennifer Erley, Sebastian Kelle, Peter Bannas, Ralf Felix Trauzeddel, Jeanette Schulz-Menger, Anja Hennemuth
Format: Article
Language:English
Published: Elsevier 2024-01-01
Series:Journal of Cardiovascular Magnetic Resonance
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1097664724011086
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850065305333661696
author Chiara Manini
Markus Hüllebrand
Lars Walczak
Sarah Nordmeyer
Lina Jarmatz
Titus Kuehne
Heiko Stern
Christian Meierhofer
Andreas Harloff
Jennifer Erley
Sebastian Kelle
Peter Bannas
Ralf Felix Trauzeddel
Jeanette Schulz-Menger
Anja Hennemuth
author_facet Chiara Manini
Markus Hüllebrand
Lars Walczak
Sarah Nordmeyer
Lina Jarmatz
Titus Kuehne
Heiko Stern
Christian Meierhofer
Andreas Harloff
Jennifer Erley
Sebastian Kelle
Peter Bannas
Ralf Felix Trauzeddel
Jeanette Schulz-Menger
Anja Hennemuth
author_sort Chiara Manini
collection DOAJ
description ABSTRACT: Background: Four-dimensional cardiovascular magnetic resonance flow imaging (4D flow CMR) plays an important role in assessing cardiovascular diseases. However, the manual or semi-automatic segmentation of aortic vessel boundaries in 4D flow data introduces variability and limits the reproducibility of aortic hemodynamics visualization and quantitative flow-related parameter computation. This paper explores the potential of deep learning to improve 4D flow CMR segmentation by developing models for automatic segmentation and analyzes the impact of the training data on the generalization of the model across different sites, scanner vendors, sequences, and pathologies. Methods: The study population consists of 260 4D flow CMR datasets, including subjects without known aortic pathology, healthy volunteers, and patients with bicuspid aortic valve (BAV) examined at different hospitals. The dataset was split to train segmentation models on subsets with different representations of characteristics, such as pathology, gender, age, scanner model, vendor, and field strength. An enhanced three-dimensional U-net convolutional neural network (CNN) architecture with residual units was trained for time-resolved two-dimensional aortic cross-sectional segmentation. Model performance was evaluated using Dice score, Hausdorff distance, and average symmetric surface distance on test data, datasets with characteristics not represented in the training set (model-specific), and an overall evaluation set. Standard diagnostic flow parameters were computed and compared with manual segmentation results using Bland-Altman analysis and interclass correlation. Results: The representation of technical factors, such as scanner vendor and field strength, in the training dataset had the strongest influence on the overall segmentation performance. Age had a greater impact than gender. Models solely trained on BAV patients’ datasets performed well on datasets of healthy subjects but not vice versa. Conclusion: This study highlights the importance of considering a heterogeneous dataset for the training of widely applicable automatic CNN segmentations in 4D flow CMR, with a particular focus on the inclusion of different pathologies and technical aspects of data acquisition.
format Article
id doaj-art-998879f0ade24441bfafde4aa4986d52
institution DOAJ
issn 1097-6647
language English
publishDate 2024-01-01
publisher Elsevier
record_format Article
series Journal of Cardiovascular Magnetic Resonance
spelling doaj-art-998879f0ade24441bfafde4aa4986d522025-08-20T02:49:02ZengElsevierJournal of Cardiovascular Magnetic Resonance1097-66472024-01-0126210108110.1016/j.jocmr.2024.101081Impact of training data composition on the generalizability of convolutional neural network aortic cross-section segmentation in four-dimensional magnetic resonance flow imagingChiara Manini0Markus Hüllebrand1Lars Walczak2Sarah Nordmeyer3Lina Jarmatz4Titus Kuehne5Heiko Stern6Christian Meierhofer7Andreas Harloff8Jennifer Erley9Sebastian Kelle10Peter Bannas11Ralf Felix Trauzeddel12Jeanette Schulz-Menger13Anja Hennemuth14Deutsches Herzzentrum der Charité (DHZC), Institute of Computer-assisted Cardiovascular Medicine, Berlin, Germany; Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt Universität zu Berlin, Berlin, Germany; Corresponding author. Deutsches Herzzentrum der Charité (DHZC), Institute of Computer-assisted Cardiovascular Medicine, Berlin, Germany.Deutsches Herzzentrum der Charité (DHZC), Institute of Computer-assisted Cardiovascular Medicine, Berlin, Germany; Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt Universität zu Berlin, Berlin, Germany; Fraunhofer MEVIS, Berlin, GermanyDeutsches Herzzentrum der Charité (DHZC), Institute of Computer-assisted Cardiovascular Medicine, Berlin, Germany; Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt Universität zu Berlin, Berlin, Germany; Fraunhofer MEVIS, Berlin, GermanyDepartment of Diagnostic and Interventional Radiology, Tübingen University Hospital, Tübingen, GermanyDeutsches Herzzentrum der Charité (DHZC), Institute of Computer-assisted Cardiovascular Medicine, Berlin, GermanyDeutsches Herzzentrum der Charité (DHZC), Institute of Computer-assisted Cardiovascular Medicine, Berlin, Germany; Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt Universität zu Berlin, Berlin, Germany; German Center for Cardiovascular Research (DZHK), Berlin, GermanyCongenital Heart Disease and Pediatric Cardiology, German Heart Center Munich, Munich, GermanyCongenital Heart Disease and Pediatric Cardiology, German Heart Center Munich, Munich, GermanyDepartment of Neurology and Neurophysiology, University Medical Center Freiburg - Faculty of Medicine, University of Freiburg, Freiburg, GermanyGerman Center for Cardiovascular Research (DZHK), Berlin, Germany; Department of Cardiology, Angiology and Intensive Care Medicine, Deutsches Herzzentrum der Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt Universität zu Berlin, Berlin, GermanyGerman Center for Cardiovascular Research (DZHK), Berlin, Germany; Department of Cardiology, Angiology and Intensive Care Medicine, Deutsches Herzzentrum der Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt Universität zu Berlin, Berlin, GermanyDepartment of Diagnostic and Interventional Radiology and Nuclear Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, GermanyGerman Center for Cardiovascular Research (DZHK), Berlin, Germany; Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, ECRC Experimental and Clinical Research Center, Lindenberger Weg 80, 13125 Berlin, Germany; Working Group on Cardiovascular Magnetic Resonance, Experimental and Clinical Research Center, a joint cooperation between the Charité – Universitätsmedizin Berlin and the Max-Delbrück-Center for Molecular Medicine, Berlin, Germany; Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Anesthesiology and Intensive Care Medicine, Charité Campus Benjamin Franklin, Hindenburgdamm 30, 12203 Berlin, GermanyGerman Center for Cardiovascular Research (DZHK), Berlin, Germany; Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, ECRC Experimental and Clinical Research Center, Lindenberger Weg 80, 13125 Berlin, Germany; Working Group on Cardiovascular Magnetic Resonance, Experimental and Clinical Research Center, a joint cooperation between the Charité – Universitätsmedizin Berlin and the Max-Delbrück-Center for Molecular Medicine, Berlin, Germany; Department of Cardiology and Nephrology, Helios Hospital Berlin-Buch, Berlin, GermanyDeutsches Herzzentrum der Charité (DHZC), Institute of Computer-assisted Cardiovascular Medicine, Berlin, Germany; Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt Universität zu Berlin, Berlin, Germany; Fraunhofer MEVIS, Berlin, Germany; German Center for Cardiovascular Research (DZHK), Berlin, Germany; Department of Diagnostic and Interventional Radiology and Nuclear Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, GermanyABSTRACT: Background: Four-dimensional cardiovascular magnetic resonance flow imaging (4D flow CMR) plays an important role in assessing cardiovascular diseases. However, the manual or semi-automatic segmentation of aortic vessel boundaries in 4D flow data introduces variability and limits the reproducibility of aortic hemodynamics visualization and quantitative flow-related parameter computation. This paper explores the potential of deep learning to improve 4D flow CMR segmentation by developing models for automatic segmentation and analyzes the impact of the training data on the generalization of the model across different sites, scanner vendors, sequences, and pathologies. Methods: The study population consists of 260 4D flow CMR datasets, including subjects without known aortic pathology, healthy volunteers, and patients with bicuspid aortic valve (BAV) examined at different hospitals. The dataset was split to train segmentation models on subsets with different representations of characteristics, such as pathology, gender, age, scanner model, vendor, and field strength. An enhanced three-dimensional U-net convolutional neural network (CNN) architecture with residual units was trained for time-resolved two-dimensional aortic cross-sectional segmentation. Model performance was evaluated using Dice score, Hausdorff distance, and average symmetric surface distance on test data, datasets with characteristics not represented in the training set (model-specific), and an overall evaluation set. Standard diagnostic flow parameters were computed and compared with manual segmentation results using Bland-Altman analysis and interclass correlation. Results: The representation of technical factors, such as scanner vendor and field strength, in the training dataset had the strongest influence on the overall segmentation performance. Age had a greater impact than gender. Models solely trained on BAV patients’ datasets performed well on datasets of healthy subjects but not vice versa. Conclusion: This study highlights the importance of considering a heterogeneous dataset for the training of widely applicable automatic CNN segmentations in 4D flow CMR, with a particular focus on the inclusion of different pathologies and technical aspects of data acquisition.http://www.sciencedirect.com/science/article/pii/S1097664724011086Deep learning4D flow MRIThoracic aortaBAVSegmentation
spellingShingle Chiara Manini
Markus Hüllebrand
Lars Walczak
Sarah Nordmeyer
Lina Jarmatz
Titus Kuehne
Heiko Stern
Christian Meierhofer
Andreas Harloff
Jennifer Erley
Sebastian Kelle
Peter Bannas
Ralf Felix Trauzeddel
Jeanette Schulz-Menger
Anja Hennemuth
Impact of training data composition on the generalizability of convolutional neural network aortic cross-section segmentation in four-dimensional magnetic resonance flow imaging
Journal of Cardiovascular Magnetic Resonance
Deep learning
4D flow MRI
Thoracic aorta
BAV
Segmentation
title Impact of training data composition on the generalizability of convolutional neural network aortic cross-section segmentation in four-dimensional magnetic resonance flow imaging
title_full Impact of training data composition on the generalizability of convolutional neural network aortic cross-section segmentation in four-dimensional magnetic resonance flow imaging
title_fullStr Impact of training data composition on the generalizability of convolutional neural network aortic cross-section segmentation in four-dimensional magnetic resonance flow imaging
title_full_unstemmed Impact of training data composition on the generalizability of convolutional neural network aortic cross-section segmentation in four-dimensional magnetic resonance flow imaging
title_short Impact of training data composition on the generalizability of convolutional neural network aortic cross-section segmentation in four-dimensional magnetic resonance flow imaging
title_sort impact of training data composition on the generalizability of convolutional neural network aortic cross section segmentation in four dimensional magnetic resonance flow imaging
topic Deep learning
4D flow MRI
Thoracic aorta
BAV
Segmentation
url http://www.sciencedirect.com/science/article/pii/S1097664724011086
work_keys_str_mv AT chiaramanini impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT markushullebrand impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT larswalczak impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT sarahnordmeyer impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT linajarmatz impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT tituskuehne impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT heikostern impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT christianmeierhofer impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT andreasharloff impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT jennifererley impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT sebastiankelle impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT peterbannas impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT ralffelixtrauzeddel impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT jeanetteschulzmenger impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging
AT anjahennemuth impactoftrainingdatacompositiononthegeneralizabilityofconvolutionalneuralnetworkaorticcrosssectionsegmentationinfourdimensionalmagneticresonanceflowimaging