Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease

Jelmer M. Wolterink; Tim Leiner; Max A. Viergever; Ivana Išgum

doi:10.1007/978-3-319-52280-7_9

Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease

Jelmer M. Wolterink
, Tim Leiner
, Max A. Viergever
, Ivana Išgum

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

24 Downloads (Pure)

Abstract

We propose an automatic method using dilated convolutional neural networks (CNNs) for segmentation of the myocardium and blood pool in cardiovascular MR (CMR) of patients with congenital heart disease (CHD). Ten training and ten test CMR scans cropped to an ROI around the heart were provided in the MICCAI 2016 HVSMR challenge. A dilated CNNwith a receptive field of 131×131 voxels was trained for myocardium and blood pool segmentation in axial, sagittal and coronal image slices. Performance was evaluated within the HVSMR challenge. Automatic segmentation of the test scans resulted in Dice indices of 0.80 ± 0.06 and 0.93 ± 0.02, average distances to boundaries of 0.96 ± 0.31 and 0.89 ± 0.24 mm, and Hausdorff distances of 6.13 ± 3.76 and 7.07 ± 3.01mm for the myocardium and blood pool, respectively. Segmentation took 41.5 ± 14.7 s per scan. In conclusion, dilated CNNs trained on a small set of CMR images of CHD patients showing large anatomical variability provide accurate myocardium and blood pool segmentations.

Original language	English
Title of host publication	Reconstruction, Segmentation, and Analysis of Medical Images - 1st International Workshops, RAMBO 2016 and HVSMR 2016 Held in Conjunction with MICCAI 2016, Revised Selected Papers
Publisher	Springer Verlag
Pages	95-102
Volume	10129 LNCS
ISBN (Print)	9783319522791
DOIs	https://doi.org/10.1007/978-3-319-52280-7_9
Publication status	Published - 2017
Event	1st International Workshops on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshops on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016 Held in Conjunction with 19th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2016 - Duration: 1 Jan 2017 → …

Publication series

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

Conference

Conference	1st International Workshops on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshops on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016 Held in Conjunction with 19th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2016
Period	01/01/2017 → …

UN SDGs

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

SDG 3 Good Health and Well-being

Access to Document

10.1007/978-3-319-52280-7_9

Dilated-convolutional-neural-networks-for-cardiovascular-mr-segmentation-in-congenital-heart-disease.pdfFinal published version, 39.9 MBLicence: Taverne

Cite this

Wolterink, J. M., Leiner, T., Viergever, M. A., & Išgum, I. (2017). Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease. In Reconstruction, Segmentation, and Analysis of Medical Images - 1st International Workshops, RAMBO 2016 and HVSMR 2016 Held in Conjunction with MICCAI 2016, Revised Selected Papers (Vol. 10129 LNCS, pp. 95-102). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Springer Verlag. https://doi.org/10.1007/978-3-319-52280-7_9

Wolterink, Jelmer M. ; Leiner, Tim ; Viergever, Max A. et al. / Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease. Reconstruction, Segmentation, and Analysis of Medical Images - 1st International Workshops, RAMBO 2016 and HVSMR 2016 Held in Conjunction with MICCAI 2016, Revised Selected Papers. Vol. 10129 LNCS Springer Verlag, 2017. pp. 95-102 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{f4519f33f9154efe85d04e27dc1cbcb8,

title = "Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease",

abstract = "We propose an automatic method using dilated convolutional neural networks (CNNs) for segmentation of the myocardium and blood pool in cardiovascular MR (CMR) of patients with congenital heart disease (CHD). Ten training and ten test CMR scans cropped to an ROI around the heart were provided in the MICCAI 2016 HVSMR challenge. A dilated CNNwith a receptive field of 131×131 voxels was trained for myocardium and blood pool segmentation in axial, sagittal and coronal image slices. Performance was evaluated within the HVSMR challenge. Automatic segmentation of the test scans resulted in Dice indices of 0.80 ± 0.06 and 0.93 ± 0.02, average distances to boundaries of 0.96 ± 0.31 and 0.89 ± 0.24 mm, and Hausdorff distances of 6.13 ± 3.76 and 7.07 ± 3.01mm for the myocardium and blood pool, respectively. Segmentation took 41.5 ± 14.7 s per scan. In conclusion, dilated CNNs trained on a small set of CMR images of CHD patients showing large anatomical variability provide accurate myocardium and blood pool segmentations.",

author = "Wolterink, \{Jelmer M.\} and Tim Leiner and Viergever, \{Max A.\} and Ivana I{\v s}gum",

year = "2017",

doi = "10.1007/978-3-319-52280-7\_9",

language = "English",

isbn = "9783319522791",

volume = "10129 LNCS",

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

publisher = "Springer Verlag",

pages = "95--102",

booktitle = "Reconstruction, Segmentation, and Analysis of Medical Images - 1st International Workshops, RAMBO 2016 and HVSMR 2016 Held in Conjunction with MICCAI 2016, Revised Selected Papers",

address = "Germany",

note = "1st International Workshops on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshops on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016 Held in Conjunction with 19th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2016 ; Conference date: 01-01-2017",

}

Wolterink, JM, Leiner, T, Viergever, MA & Išgum, I 2017, Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease. in Reconstruction, Segmentation, and Analysis of Medical Images - 1st International Workshops, RAMBO 2016 and HVSMR 2016 Held in Conjunction with MICCAI 2016, Revised Selected Papers. vol. 10129 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Springer Verlag, pp. 95-102, 1st International Workshops on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshops on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016 Held in Conjunction with 19th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2016, 01/01/2017. https://doi.org/10.1007/978-3-319-52280-7_9

Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease. / Wolterink, Jelmer M.; Leiner, Tim; Viergever, Max A. et al.
Reconstruction, Segmentation, and Analysis of Medical Images - 1st International Workshops, RAMBO 2016 and HVSMR 2016 Held in Conjunction with MICCAI 2016, Revised Selected Papers. Vol. 10129 LNCS Springer Verlag, 2017. p. 95-102 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

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

TY - GEN

T1 - Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease

AU - Wolterink, Jelmer M.

AU - Leiner, Tim

AU - Viergever, Max A.

AU - Išgum, Ivana

PY - 2017

Y1 - 2017

N2 - We propose an automatic method using dilated convolutional neural networks (CNNs) for segmentation of the myocardium and blood pool in cardiovascular MR (CMR) of patients with congenital heart disease (CHD). Ten training and ten test CMR scans cropped to an ROI around the heart were provided in the MICCAI 2016 HVSMR challenge. A dilated CNNwith a receptive field of 131×131 voxels was trained for myocardium and blood pool segmentation in axial, sagittal and coronal image slices. Performance was evaluated within the HVSMR challenge. Automatic segmentation of the test scans resulted in Dice indices of 0.80 ± 0.06 and 0.93 ± 0.02, average distances to boundaries of 0.96 ± 0.31 and 0.89 ± 0.24 mm, and Hausdorff distances of 6.13 ± 3.76 and 7.07 ± 3.01mm for the myocardium and blood pool, respectively. Segmentation took 41.5 ± 14.7 s per scan. In conclusion, dilated CNNs trained on a small set of CMR images of CHD patients showing large anatomical variability provide accurate myocardium and blood pool segmentations.

AB - We propose an automatic method using dilated convolutional neural networks (CNNs) for segmentation of the myocardium and blood pool in cardiovascular MR (CMR) of patients with congenital heart disease (CHD). Ten training and ten test CMR scans cropped to an ROI around the heart were provided in the MICCAI 2016 HVSMR challenge. A dilated CNNwith a receptive field of 131×131 voxels was trained for myocardium and blood pool segmentation in axial, sagittal and coronal image slices. Performance was evaluated within the HVSMR challenge. Automatic segmentation of the test scans resulted in Dice indices of 0.80 ± 0.06 and 0.93 ± 0.02, average distances to boundaries of 0.96 ± 0.31 and 0.89 ± 0.24 mm, and Hausdorff distances of 6.13 ± 3.76 and 7.07 ± 3.01mm for the myocardium and blood pool, respectively. Segmentation took 41.5 ± 14.7 s per scan. In conclusion, dilated CNNs trained on a small set of CMR images of CHD patients showing large anatomical variability provide accurate myocardium and blood pool segmentations.

UR - https://www.scopus.com/pages/publications/85011307871

U2 - 10.1007/978-3-319-52280-7_9

DO - 10.1007/978-3-319-52280-7_9

M3 - Conference contribution

SN - 9783319522791

VL - 10129 LNCS

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

SP - 95

EP - 102

BT - Reconstruction, Segmentation, and Analysis of Medical Images - 1st International Workshops, RAMBO 2016 and HVSMR 2016 Held in Conjunction with MICCAI 2016, Revised Selected Papers

PB - Springer Verlag

T2 - 1st International Workshops on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshops on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016 Held in Conjunction with 19th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2016

Y2 - 1 January 2017

ER -

Wolterink JM, Leiner T, Viergever MA, Išgum I. Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease. In Reconstruction, Segmentation, and Analysis of Medical Images - 1st International Workshops, RAMBO 2016 and HVSMR 2016 Held in Conjunction with MICCAI 2016, Revised Selected Papers. Vol. 10129 LNCS. Springer Verlag. 2017. p. 95-102. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-52280-7_9

Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease

Abstract

Publication series

Conference

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this