Shortcomings of ventricle segmentation using deep convolutional networks

Muhan Shao, Shuo Han, Aaron Carass, Xiang Li, Ari M Blitz, Jerry Ladd Prince, Lotta M. Ellingsen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Normal Pressure Hydrocephalus (NPH) is a brain disorder that can present with ventriculomegaly and dementia-like symptoms, which often can be reversed through surgery. Having accurate segmentation of the ventricular system into its sub-compartments from magnetic resonance images (MRI) would be beneficial to better characterize the condition of NPH patients. Previous segmentation algorithms need long processing time and often fail to accurately segment severely enlarged ventricles in NPH patients. Recently, deep convolutional neural network (CNN) methods have been reported to have fast and accurate performance on medical image segmentation tasks. In this paper, we present a 3D U-net CNN-based network to segment the ventricular system in MRI. We trained three networks on different data sets and compared their performances. The networks trained on healthy controls (HC) failed in patients with NPH pathology, even in patients with normal appearing ventricles. The network trained on images from HC and NPH patients provided superior performance against state-of-the-art methods when evaluated on images from both data sets.

Original languageEnglish (US)
Title of host publicationUnderstanding and Interpreting Machine Learning in Medical Image Computing Applications - First International Workshops MLCN 2018, DLF 2018, and iMIMIC 2018, Held in Conjunction with MICCAI 2018, Proceedings
EditorsZeike Taylor, Mauricio Reyes, M. Jorge Cardoso, Carlos A. Silva, Danail Stoyanov, Lena Maier-Hein, Sergio Pereira, Seyed Mostafa Kia, Ipek Oguz, Bennett Landman, Anne Martel, Edouard Duchesnay, Tommy Lofstedt, Andre F. Marquand, Raphael Meier
PublisherSpringer Verlag
Pages79-86
Number of pages8
ISBN (Print)9783030026271
DOIs
StatePublished - Jan 1 2018
Event1st International Workshop on Machine Learning in Clinical Neuroimaging, MLCN 2018, 1st International Workshop on Deep Learning Fails, DLF 2018, and 1st International Workshop on Interpretability of Machine Intelligence in Medical Image Computing, iMIMIC 2018, held in conjunction with the 21st International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2018 - Granada, Spain
Duration: Sep 16 2018Sep 20 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11038 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other1st International Workshop on Machine Learning in Clinical Neuroimaging, MLCN 2018, 1st International Workshop on Deep Learning Fails, DLF 2018, and 1st International Workshop on Interpretability of Machine Intelligence in Medical Image Computing, iMIMIC 2018, held in conjunction with the 21st International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2018
CountrySpain
CityGranada
Period9/16/189/20/18

Fingerprint

Segmentation
Magnetic Resonance Image
Magnetic resonance
Neural Networks
Neural networks
Dementia
Pathology
Medical Image
Image segmentation
Image Segmentation
Surgery
Disorder
Brain
Processing

Keywords

  • CNN
  • Hydrocephalus
  • MRI
  • Segmentation

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Shao, M., Han, S., Carass, A., Li, X., Blitz, A. M., Prince, J. L., & Ellingsen, L. M. (2018). Shortcomings of ventricle segmentation using deep convolutional networks. In Z. Taylor, M. Reyes, M. J. Cardoso, C. A. Silva, D. Stoyanov, L. Maier-Hein, S. Pereira, S. M. Kia, I. Oguz, B. Landman, A. Martel, E. Duchesnay, T. Lofstedt, A. F. Marquand, ... R. Meier (Eds.), Understanding and Interpreting Machine Learning in Medical Image Computing Applications - First International Workshops MLCN 2018, DLF 2018, and iMIMIC 2018, Held in Conjunction with MICCAI 2018, Proceedings (pp. 79-86). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11038 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-02628-8_9

Shortcomings of ventricle segmentation using deep convolutional networks. / Shao, Muhan; Han, Shuo; Carass, Aaron; Li, Xiang; Blitz, Ari M; Prince, Jerry Ladd; Ellingsen, Lotta M.

Understanding and Interpreting Machine Learning in Medical Image Computing Applications - First International Workshops MLCN 2018, DLF 2018, and iMIMIC 2018, Held in Conjunction with MICCAI 2018, Proceedings. ed. / Zeike Taylor; Mauricio Reyes; M. Jorge Cardoso; Carlos A. Silva; Danail Stoyanov; Lena Maier-Hein; Sergio Pereira; Seyed Mostafa Kia; Ipek Oguz; Bennett Landman; Anne Martel; Edouard Duchesnay; Tommy Lofstedt; Andre F. Marquand; Raphael Meier. Springer Verlag, 2018. p. 79-86 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11038 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Shao, M, Han, S, Carass, A, Li, X, Blitz, AM, Prince, JL & Ellingsen, LM 2018, Shortcomings of ventricle segmentation using deep convolutional networks. in Z Taylor, M Reyes, MJ Cardoso, CA Silva, D Stoyanov, L Maier-Hein, S Pereira, SM Kia, I Oguz, B Landman, A Martel, E Duchesnay, T Lofstedt, AF Marquand & R Meier (eds), Understanding and Interpreting Machine Learning in Medical Image Computing Applications - First International Workshops MLCN 2018, DLF 2018, and iMIMIC 2018, Held in Conjunction with MICCAI 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11038 LNCS, Springer Verlag, pp. 79-86, 1st International Workshop on Machine Learning in Clinical Neuroimaging, MLCN 2018, 1st International Workshop on Deep Learning Fails, DLF 2018, and 1st International Workshop on Interpretability of Machine Intelligence in Medical Image Computing, iMIMIC 2018, held in conjunction with the 21st International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2018, Granada, Spain, 9/16/18. https://doi.org/10.1007/978-3-030-02628-8_9
Shao M, Han S, Carass A, Li X, Blitz AM, Prince JL et al. Shortcomings of ventricle segmentation using deep convolutional networks. In Taylor Z, Reyes M, Cardoso MJ, Silva CA, Stoyanov D, Maier-Hein L, Pereira S, Kia SM, Oguz I, Landman B, Martel A, Duchesnay E, Lofstedt T, Marquand AF, Meier R, editors, Understanding and Interpreting Machine Learning in Medical Image Computing Applications - First International Workshops MLCN 2018, DLF 2018, and iMIMIC 2018, Held in Conjunction with MICCAI 2018, Proceedings. Springer Verlag. 2018. p. 79-86. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-02628-8_9
Shao, Muhan ; Han, Shuo ; Carass, Aaron ; Li, Xiang ; Blitz, Ari M ; Prince, Jerry Ladd ; Ellingsen, Lotta M. / Shortcomings of ventricle segmentation using deep convolutional networks. Understanding and Interpreting Machine Learning in Medical Image Computing Applications - First International Workshops MLCN 2018, DLF 2018, and iMIMIC 2018, Held in Conjunction with MICCAI 2018, Proceedings. editor / Zeike Taylor ; Mauricio Reyes ; M. Jorge Cardoso ; Carlos A. Silva ; Danail Stoyanov ; Lena Maier-Hein ; Sergio Pereira ; Seyed Mostafa Kia ; Ipek Oguz ; Bennett Landman ; Anne Martel ; Edouard Duchesnay ; Tommy Lofstedt ; Andre F. Marquand ; Raphael Meier. Springer Verlag, 2018. pp. 79-86 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{04f7c399c80644479c85845afc02de48,
title = "Shortcomings of ventricle segmentation using deep convolutional networks",
abstract = "Normal Pressure Hydrocephalus (NPH) is a brain disorder that can present with ventriculomegaly and dementia-like symptoms, which often can be reversed through surgery. Having accurate segmentation of the ventricular system into its sub-compartments from magnetic resonance images (MRI) would be beneficial to better characterize the condition of NPH patients. Previous segmentation algorithms need long processing time and often fail to accurately segment severely enlarged ventricles in NPH patients. Recently, deep convolutional neural network (CNN) methods have been reported to have fast and accurate performance on medical image segmentation tasks. In this paper, we present a 3D U-net CNN-based network to segment the ventricular system in MRI. We trained three networks on different data sets and compared their performances. The networks trained on healthy controls (HC) failed in patients with NPH pathology, even in patients with normal appearing ventricles. The network trained on images from HC and NPH patients provided superior performance against state-of-the-art methods when evaluated on images from both data sets.",
keywords = "CNN, Hydrocephalus, MRI, Segmentation",
author = "Muhan Shao and Shuo Han and Aaron Carass and Xiang Li and Blitz, {Ari M} and Prince, {Jerry Ladd} and Ellingsen, {Lotta M.}",
year = "2018",
month = "1",
day = "1",
doi = "10.1007/978-3-030-02628-8_9",
language = "English (US)",
isbn = "9783030026271",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "79--86",
editor = "Zeike Taylor and Mauricio Reyes and Cardoso, {M. Jorge} and Silva, {Carlos A.} and Danail Stoyanov and Lena Maier-Hein and Sergio Pereira and Kia, {Seyed Mostafa} and Ipek Oguz and Bennett Landman and Anne Martel and Edouard Duchesnay and Tommy Lofstedt and Marquand, {Andre F.} and Raphael Meier",
booktitle = "Understanding and Interpreting Machine Learning in Medical Image Computing Applications - First International Workshops MLCN 2018, DLF 2018, and iMIMIC 2018, Held in Conjunction with MICCAI 2018, Proceedings",

}

TY - GEN

T1 - Shortcomings of ventricle segmentation using deep convolutional networks

AU - Shao, Muhan

AU - Han, Shuo

AU - Carass, Aaron

AU - Li, Xiang

AU - Blitz, Ari M

AU - Prince, Jerry Ladd

AU - Ellingsen, Lotta M.

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Normal Pressure Hydrocephalus (NPH) is a brain disorder that can present with ventriculomegaly and dementia-like symptoms, which often can be reversed through surgery. Having accurate segmentation of the ventricular system into its sub-compartments from magnetic resonance images (MRI) would be beneficial to better characterize the condition of NPH patients. Previous segmentation algorithms need long processing time and often fail to accurately segment severely enlarged ventricles in NPH patients. Recently, deep convolutional neural network (CNN) methods have been reported to have fast and accurate performance on medical image segmentation tasks. In this paper, we present a 3D U-net CNN-based network to segment the ventricular system in MRI. We trained three networks on different data sets and compared their performances. The networks trained on healthy controls (HC) failed in patients with NPH pathology, even in patients with normal appearing ventricles. The network trained on images from HC and NPH patients provided superior performance against state-of-the-art methods when evaluated on images from both data sets.

AB - Normal Pressure Hydrocephalus (NPH) is a brain disorder that can present with ventriculomegaly and dementia-like symptoms, which often can be reversed through surgery. Having accurate segmentation of the ventricular system into its sub-compartments from magnetic resonance images (MRI) would be beneficial to better characterize the condition of NPH patients. Previous segmentation algorithms need long processing time and often fail to accurately segment severely enlarged ventricles in NPH patients. Recently, deep convolutional neural network (CNN) methods have been reported to have fast and accurate performance on medical image segmentation tasks. In this paper, we present a 3D U-net CNN-based network to segment the ventricular system in MRI. We trained three networks on different data sets and compared their performances. The networks trained on healthy controls (HC) failed in patients with NPH pathology, even in patients with normal appearing ventricles. The network trained on images from HC and NPH patients provided superior performance against state-of-the-art methods when evaluated on images from both data sets.

KW - CNN

KW - Hydrocephalus

KW - MRI

KW - Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85056458906&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85056458906&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-02628-8_9

DO - 10.1007/978-3-030-02628-8_9

M3 - Conference contribution

AN - SCOPUS:85056458906

SN - 9783030026271

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

SP - 79

EP - 86

BT - Understanding and Interpreting Machine Learning in Medical Image Computing Applications - First International Workshops MLCN 2018, DLF 2018, and iMIMIC 2018, Held in Conjunction with MICCAI 2018, Proceedings

A2 - Taylor, Zeike

A2 - Reyes, Mauricio

A2 - Cardoso, M. Jorge

A2 - Silva, Carlos A.

A2 - Stoyanov, Danail

A2 - Maier-Hein, Lena

A2 - Pereira, Sergio

A2 - Kia, Seyed Mostafa

A2 - Oguz, Ipek

A2 - Landman, Bennett

A2 - Martel, Anne

A2 - Duchesnay, Edouard

A2 - Lofstedt, Tommy

A2 - Marquand, Andre F.

A2 - Meier, Raphael

PB - Springer Verlag

ER -