Publication detail

Segmentation of Head and Neck Organs at Risk Using CNN with Batch Dice Loss

KODYM, O. ŠPANĚL, M. HEROUT, A.

Original Title

Segmentation of Head and Neck Organs at Risk Using CNN with Batch Dice Loss

English Title

Segmentation of Head and Neck Organs at Risk Using CNN with Batch Dice Loss

Type

conference paper

Language

en

Original Abstract

This paper deals with segmentation of organs at risk (OAR) in head and neck area in CT images which is a crucial step for reliable intensity modulated radiotherapy treatment. We introduce a convolution neural network with encoder-decoder architecture and a new loss function, the batch soft Dice loss function, used to train the network. The resulting model produces segmentations of every OAR in the public MICCAI 2015 Head And Neck Auto-Segmentation Challenge dataset. Despite the heavy class imbalance in the data, we improve accuracy of current state-of-the-art methods by 0.33 mm in terms of average surface distance and by 0.11 in terms of Dice overlap coefficient on average. 

English abstract

This paper deals with segmentation of organs at risk (OAR) in head and neck area in CT images which is a crucial step for reliable intensity modulated radiotherapy treatment. We introduce a convolution neural network with encoder-decoder architecture and a new loss function, the batch soft Dice loss function, used to train the network. The resulting model produces segmentations of every OAR in the public MICCAI 2015 Head And Neck Auto-Segmentation Challenge dataset. Despite the heavy class imbalance in the data, we improve accuracy of current state-of-the-art methods by 0.33 mm in terms of average surface distance and by 0.11 in terms of Dice overlap coefficient on average. 

Keywords

Convolutional Neural Networks, Computed Tomography, Multi-label Segmentation, Head and Neck Radiotherapy

Released

20.07.2018

Publisher

Springer International Publishing

Location

Stuttgart

ISBN

978-3-030-12938-5

Book

Pattern Recognition, 40th German Conference, GCPR 2018 Proceedings

Edition

LNCS, volume 11269

Edition number

NEUVEDEN

Pages from

105

Pages to

114

Pages count

9

Documents

BibTex


@inproceedings{BUT155013,
  author="Oldřich {Kodym} and Michal {Španěl} and Adam {Herout}",
  title="Segmentation of Head and Neck Organs at Risk Using CNN with Batch Dice Loss",
  annote="This paper deals with segmentation of organs at risk (OAR) in head and neck area
in CT images which is a crucial step for reliable intensity modulated
radiotherapy treatment. We introduce a convolution neural network with
encoder-decoder architecture and a new loss function, the batch soft Dice loss
function, used to train the network. The resulting model produces segmentations
of every OAR in the public MICCAI 2015 Head And Neck Auto-Segmentation Challenge
dataset. Despite the heavy class imbalance in the data, we improve accuracy of
current state-of-the-art methods by 0.33 mm in terms of average surface distance
and by 0.11 in terms of Dice overlap coefficient on average. ",
  address="Springer International Publishing",
  booktitle="Pattern Recognition, 40th German Conference, GCPR 2018 Proceedings",
  chapter="155013",
  doi="10.1007/978-3-030-12939-2_8",
  edition="LNCS, volume 11269",
  howpublished="online",
  institution="Springer International Publishing",
  number="11269",
  year="2018",
  month="july",
  pages="105--114",
  publisher="Springer International Publishing",
  type="conference paper"
}