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

^*Corresponding author for this work

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

59 Citations (Scopus)

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
Subtitle of host publication	First International Workshops, RAMBO 2016 and HVSMR 2016, Held in Conjunction with MICCAI 2016, Athens, Greece, October 17, 2016, Revised Selected Papers
Editors	Maria A. Zuluaga, Mehdi H. Moghari, Danielle F. Pace, Bernhard Kainz, Kanwal Bhatia
Place of Publication	Cham
Publisher	Springer
Pages	95-102
Number of pages	8
ISBN (Electronic)	978-3-319-52280-7
ISBN (Print)	978-3-319-52279-1
DOIs	https://doi.org/10.1007/978-3-319-52280-7_9
Publication status	Published - 1 Jan 2017
Externally published	Yes
Event	1st International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshop on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016 - Athens, Greece Duration: 17 Oct 2016 → 21 Oct 2016 Conference number: 1

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	10129
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	1st International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshop on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016
Abbreviated title	RAMBOI-HVSMR
Country/Territory	Greece
City	Athens
Period	17/10/16 → 21/10/16

Keywords

Cardiovascular MR
Congenital heart disease
Deep learning
Dilated convolutional neural networks
Medical image segmentation

UN SDGs

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

Access to Document

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

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 M. A. Zuluaga, M. H. Moghari, D. F. Pace, B. Kainz, & K. Bhatia (Eds.), Reconstruction, Segmentation, and Analysis of Medical Images: First International Workshops, RAMBO 2016 and HVSMR 2016, Held in Conjunction with MICCAI 2016, Athens, Greece, October 17, 2016, Revised Selected Papers (pp. 95-102). (Lecture Notes in Computer Science; Vol. 10129). Springer. 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: First International Workshops, RAMBO 2016 and HVSMR 2016, Held in Conjunction with MICCAI 2016, Athens, Greece, October 17, 2016, Revised Selected Papers. editor / Maria A. Zuluaga ; Mehdi H. Moghari ; Danielle F. Pace ; Bernhard Kainz ; Kanwal Bhatia. Cham : Springer, 2017. pp. 95-102 (Lecture Notes in Computer Science).

@inproceedings{4665b1dd01224d1c9dd7f1952311f933,

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.",

keywords = "Cardiovascular MR, Congenital heart disease, Deep learning, Dilated convolutional neural networks, Medical image segmentation",

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

year = "2017",

month = jan,

day = "1",

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

language = "English",

isbn = "978-3-319-52279-1",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "95--102",

editor = "Zuluaga, {Maria A.} and Moghari, {Mehdi H.} and Pace, {Danielle F.} and Bernhard Kainz and Kanwal Bhatia",

booktitle = "Reconstruction, Segmentation, and Analysis of Medical Images",

address = "Germany",

note = "1st International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshop on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016, RAMBOI-HVSMR ; Conference date: 17-10-2016 Through 21-10-2016",

}

Wolterink, JM, Leiner, T, Viergever, MA & Išgum, I 2017, Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease. in MA Zuluaga, MH Moghari, DF Pace, B Kainz & K Bhatia (eds), Reconstruction, Segmentation, and Analysis of Medical Images: First International Workshops, RAMBO 2016 and HVSMR 2016, Held in Conjunction with MICCAI 2016, Athens, Greece, October 17, 2016, Revised Selected Papers. Lecture Notes in Computer Science, vol. 10129, Springer, Cham, pp. 95-102, 1st International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshop on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016, Athens, Greece, 17/10/16. 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: First International Workshops, RAMBO 2016 and HVSMR 2016, Held in Conjunction with MICCAI 2016, Athens, Greece, October 17, 2016, Revised Selected Papers. ed. / Maria A. Zuluaga; Mehdi H. Moghari; Danielle F. Pace; Bernhard Kainz; Kanwal Bhatia. Cham: Springer, 2017. p. 95-102 (Lecture Notes in Computer Science; Vol. 10129).

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

N1 - Conference code: 1

PY - 2017/1/1

Y1 - 2017/1/1

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.

KW - Cardiovascular MR

KW - Congenital heart disease

KW - Deep learning

KW - Dilated convolutional neural networks

KW - Medical image segmentation

UR - http://www.scopus.com/inward/record.url?scp=85011307871&partnerID=8YFLogxK

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

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

M3 - Conference contribution

AN - SCOPUS:85011307871

SN - 978-3-319-52279-1

T3 - Lecture Notes in Computer Science

SP - 95

EP - 102

BT - Reconstruction, Segmentation, and Analysis of Medical Images

A2 - Zuluaga, Maria A.

A2 - Moghari, Mehdi H.

A2 - Pace, Danielle F.

A2 - Kainz, Bernhard

A2 - Bhatia, Kanwal

PB - Springer

CY - Cham

T2 - 1st International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2016 and 1st International Workshop on Whole-Heart and Great Vessel Segmentation from 3D Cardiovascular MRI in Congenital Heart Disease, HVSMR 2016

Y2 - 17 October 2016 through 21 October 2016

ER -

Wolterink JM, Leiner T, Viergever MA, Išgum I. Dilated convolutional neural networks for cardiovascular MR segmentation in congenital heart disease. In Zuluaga MA, Moghari MH, Pace DF, Kainz B, Bhatia K, editors, Reconstruction, Segmentation, and Analysis of Medical Images: First International Workshops, RAMBO 2016 and HVSMR 2016, Held in Conjunction with MICCAI 2016, Athens, Greece, October 17, 2016, Revised Selected Papers. Cham: Springer. 2017. p. 95-102. (Lecture Notes in Computer Science). 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

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this