Segmentation of head and neck tumours using modified U-net

Research output: Contribution to conferencePaper

Abstract

A new neural network for automatic head and neck cancer (HNC) segmentation from magnetic resonance imaging (MRI) is presented. The proposed neural network is based on U-net, which combines features from different resolutions to achieve end-to-end locating and segmentation of medical images. In this work, the dilated convolution is introduced into U-net, to obtain larger receptive field so that extract multi-scale features. Also, this network uses Dice loss to reduce the imbalance between classes. The proposed algorithm is trained and tested on real MRI data. The cross-validation results show that the new network outperformed the original Unet by 5% (Dice score) on head and neck tumour segmentation.

Conference

Conference27th European Signal Processing Conference
Abbreviated titleEUSIPCO
CountrySpain
CityA Coruna
Period2/09/196/09/19
Internet address

Fingerprint

Magnetic resonance
Tumors
Neck
Head
Magnetic Resonance Imaging
Neural networks
Imaging techniques
Head and Neck Neoplasms
Convolution
Neoplasms

Keywords

  • MRI data
  • head and neck cancer
  • U-net
  • dilated convolution
  • semantic segmentation

Cite this

Zhao, B., Soraghan, J., Di Caterina, G., & Grose, D. (2019). Segmentation of head and neck tumours using modified U-net. Paper presented at 27th European Signal Processing Conference, A Coruna, Spain.
Zhao, Baixiang ; Soraghan, John ; Di Caterina, Gaetano ; Grose, Derek. / Segmentation of head and neck tumours using modified U-net. Paper presented at 27th European Signal Processing Conference, A Coruna, Spain.4 p.
@conference{a04ff4daaf28450c9b3d84bdea8c7727,
title = "Segmentation of head and neck tumours using modified U-net",
abstract = "A new neural network for automatic head and neck cancer (HNC) segmentation from magnetic resonance imaging (MRI) is presented. The proposed neural network is based on U-net, which combines features from different resolutions to achieve end-to-end locating and segmentation of medical images. In this work, the dilated convolution is introduced into U-net, to obtain larger receptive field so that extract multi-scale features. Also, this network uses Dice loss to reduce the imbalance between classes. The proposed algorithm is trained and tested on real MRI data. The cross-validation results show that the new network outperformed the original Unet by 5{\%} (Dice score) on head and neck tumour segmentation.",
keywords = "MRI data, head and neck cancer, U-net, dilated convolution, semantic segmentation",
author = "Baixiang Zhao and John Soraghan and {Di Caterina}, Gaetano and Derek Grose",
year = "2019",
month = "9",
day = "2",
language = "English",
note = "27th European Signal Processing Conference, EUSIPCO ; Conference date: 02-09-2019 Through 06-09-2019",
url = "http://eusipco2019.org/",

}

Zhao, B, Soraghan, J, Di Caterina, G & Grose, D 2019, 'Segmentation of head and neck tumours using modified U-net' Paper presented at 27th European Signal Processing Conference, A Coruna, Spain, 2/09/19 - 6/09/19, .

Segmentation of head and neck tumours using modified U-net. / Zhao, Baixiang; Soraghan, John; Di Caterina, Gaetano; Grose, Derek.

2019. Paper presented at 27th European Signal Processing Conference, A Coruna, Spain.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Segmentation of head and neck tumours using modified U-net

AU - Zhao, Baixiang

AU - Soraghan, John

AU - Di Caterina, Gaetano

AU - Grose, Derek

PY - 2019/9/2

Y1 - 2019/9/2

N2 - A new neural network for automatic head and neck cancer (HNC) segmentation from magnetic resonance imaging (MRI) is presented. The proposed neural network is based on U-net, which combines features from different resolutions to achieve end-to-end locating and segmentation of medical images. In this work, the dilated convolution is introduced into U-net, to obtain larger receptive field so that extract multi-scale features. Also, this network uses Dice loss to reduce the imbalance between classes. The proposed algorithm is trained and tested on real MRI data. The cross-validation results show that the new network outperformed the original Unet by 5% (Dice score) on head and neck tumour segmentation.

AB - A new neural network for automatic head and neck cancer (HNC) segmentation from magnetic resonance imaging (MRI) is presented. The proposed neural network is based on U-net, which combines features from different resolutions to achieve end-to-end locating and segmentation of medical images. In this work, the dilated convolution is introduced into U-net, to obtain larger receptive field so that extract multi-scale features. Also, this network uses Dice loss to reduce the imbalance between classes. The proposed algorithm is trained and tested on real MRI data. The cross-validation results show that the new network outperformed the original Unet by 5% (Dice score) on head and neck tumour segmentation.

KW - MRI data

KW - head and neck cancer

KW - U-net

KW - dilated convolution

KW - semantic segmentation

UR - http://eusipco2019.org/

M3 - Paper

ER -

Zhao B, Soraghan J, Di Caterina G, Grose D. Segmentation of head and neck tumours using modified U-net. 2019. Paper presented at 27th European Signal Processing Conference, A Coruna, Spain.