Fine-grain segmentation of the intervertebral discs from MR spine images using deep convolutional neural networks: BSU-Net

Sewon Kim, Won C. Bae, Koichi Masuda, Christine B. Chung, Do Sik Hwang

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

We propose a new deep learning network capable of successfully segmenting intervertebral discs and their complex boundaries from magnetic resonance (MR) spine images. The existing U-network (U-net) is known to perform well in various segmentation tasks in medical images; however, its performance with respect to details of segmentation such as boundaries is limited by the structural limitations of a max-pooling layer that plays a key role in feature extraction process in the U-net. We designed a modified convolutional and pooling layer scheme and applied a cascaded learning method to overcome these structural limitations of the max-pooling layer of a conventional U-net. The proposed network achieved 3% higher Dice similarity coefficient (DSC) than conventional U-net for intervertebral disc segmentation (89.44% vs. 86.44%, respectively; p < 0.001). For intervertebral disc boundary segmentation, the proposed network achieved 10.46% higher DSC than conventional U-net (54.62% vs. 44.16%, respectively; p < 0.001).

Original languageEnglish
Article number1656
JournalApplied Sciences (Switzerland)
Volume8
Issue number9
DOIs
Publication statusPublished - 2018 Sep 14

Fingerprint

intervertebral disks
spine
Magnetic resonance
magnetic resonance
Feature extraction
Neural networks
learning
coefficients
pattern recognition
Deep learning

All Science Journal Classification (ASJC) codes

  • Materials Science(all)
  • Instrumentation
  • Engineering(all)
  • Process Chemistry and Technology
  • Computer Science Applications
  • Fluid Flow and Transfer Processes

Cite this

@article{13b5a9dbcac244d49a8df87057fe1572,
title = "Fine-grain segmentation of the intervertebral discs from MR spine images using deep convolutional neural networks: BSU-Net",
abstract = "We propose a new deep learning network capable of successfully segmenting intervertebral discs and their complex boundaries from magnetic resonance (MR) spine images. The existing U-network (U-net) is known to perform well in various segmentation tasks in medical images; however, its performance with respect to details of segmentation such as boundaries is limited by the structural limitations of a max-pooling layer that plays a key role in feature extraction process in the U-net. We designed a modified convolutional and pooling layer scheme and applied a cascaded learning method to overcome these structural limitations of the max-pooling layer of a conventional U-net. The proposed network achieved 3{\%} higher Dice similarity coefficient (DSC) than conventional U-net for intervertebral disc segmentation (89.44{\%} vs. 86.44{\%}, respectively; p < 0.001). For intervertebral disc boundary segmentation, the proposed network achieved 10.46{\%} higher DSC than conventional U-net (54.62{\%} vs. 44.16{\%}, respectively; p < 0.001).",
author = "Sewon Kim and Bae, {Won C.} and Koichi Masuda and Chung, {Christine B.} and Hwang, {Do Sik}",
year = "2018",
month = "9",
day = "14",
doi = "10.3390/app8091656",
language = "English",
volume = "8",
journal = "Applied Sciences (Switzerland)",
issn = "2076-3417",
publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",
number = "9",

}

Fine-grain segmentation of the intervertebral discs from MR spine images using deep convolutional neural networks : BSU-Net. / Kim, Sewon; Bae, Won C.; Masuda, Koichi; Chung, Christine B.; Hwang, Do Sik.

In: Applied Sciences (Switzerland), Vol. 8, No. 9, 1656, 14.09.2018.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Fine-grain segmentation of the intervertebral discs from MR spine images using deep convolutional neural networks

T2 - BSU-Net

AU - Kim, Sewon

AU - Bae, Won C.

AU - Masuda, Koichi

AU - Chung, Christine B.

AU - Hwang, Do Sik

PY - 2018/9/14

Y1 - 2018/9/14

N2 - We propose a new deep learning network capable of successfully segmenting intervertebral discs and their complex boundaries from magnetic resonance (MR) spine images. The existing U-network (U-net) is known to perform well in various segmentation tasks in medical images; however, its performance with respect to details of segmentation such as boundaries is limited by the structural limitations of a max-pooling layer that plays a key role in feature extraction process in the U-net. We designed a modified convolutional and pooling layer scheme and applied a cascaded learning method to overcome these structural limitations of the max-pooling layer of a conventional U-net. The proposed network achieved 3% higher Dice similarity coefficient (DSC) than conventional U-net for intervertebral disc segmentation (89.44% vs. 86.44%, respectively; p < 0.001). For intervertebral disc boundary segmentation, the proposed network achieved 10.46% higher DSC than conventional U-net (54.62% vs. 44.16%, respectively; p < 0.001).

AB - We propose a new deep learning network capable of successfully segmenting intervertebral discs and their complex boundaries from magnetic resonance (MR) spine images. The existing U-network (U-net) is known to perform well in various segmentation tasks in medical images; however, its performance with respect to details of segmentation such as boundaries is limited by the structural limitations of a max-pooling layer that plays a key role in feature extraction process in the U-net. We designed a modified convolutional and pooling layer scheme and applied a cascaded learning method to overcome these structural limitations of the max-pooling layer of a conventional U-net. The proposed network achieved 3% higher Dice similarity coefficient (DSC) than conventional U-net for intervertebral disc segmentation (89.44% vs. 86.44%, respectively; p < 0.001). For intervertebral disc boundary segmentation, the proposed network achieved 10.46% higher DSC than conventional U-net (54.62% vs. 44.16%, respectively; p < 0.001).

UR - http://www.scopus.com/inward/record.url?scp=85053425285&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053425285&partnerID=8YFLogxK

U2 - 10.3390/app8091656

DO - 10.3390/app8091656

M3 - Article

AN - SCOPUS:85053425285

VL - 8

JO - Applied Sciences (Switzerland)

JF - Applied Sciences (Switzerland)

SN - 2076-3417

IS - 9

M1 - 1656

ER -