Quantification of Lung Abnormalities in Cystic Fibrosis using Deep Networks

Filipe Marques; Florian Dubost; Mariette Kemner-van de Corput; Harm A. W. Tiddens; Marleen de Bruijne

doi:10.1117/12.2292188

Quantification of Lung Abnormalities in Cystic Fibrosis using Deep Networks

Filipe Marques, Florian Dubost, Mariette Kemner-van de Corput, Harm A. W. Tiddens, Marleen de Bruijne

3 Citations (Scopus)

Abstract

Cystic fibrosis is a genetic disease which may appear in early life with structural abnormalities in lung tissues. We propose to detect these abnormalities using a texture classification approach. Our method is a cascade of two convolutional neural networks. The first network detects the presence of abnormal tissues. The second network identifies the type of the structural abnormalities: bronchiectasis, atelectasis or mucus plugging.We also propose a network computing pixel-wise heatmaps of abnormality presence learning only from the patch-wise annotations. Our database consists of CT scans of 194 subjects. We use 154 subjects to train our algorithms and the 40 remaining ones as a test set. We compare our method with random forest and a single neural network approach. The first network reaches a sensitivity of 0,62 for disease detection, 0,10 higher than the random forest classifier and 0,17 higher than the single neural network. Our cascade approach yields a final class-averaged F1-score of 0,38, outperforming the baseline method and the single network by 0,15 and 0,10.

Original language	English
Title of host publication	Medical Imaging 2018 : Image Processing
Number of pages	7
Publisher	SPIE - International Society for Optical Engineering
Publication date	2018
Article number	105741G
DOIs	https://doi.org/10.1117/12.2292188
Publication status	Published - 2018
Event	SPIE Medical Imaging 2018 - Houston, United States Duration: 10 Feb 2018 → 15 Feb 2018

Conference

Conference	SPIE Medical Imaging 2018
Country/Territory	United States
City	Houston
Period	10/02/2018 → 15/02/2018

Series	Proceedings of SPIE International Symposium on Medical Imaging
Volume	10574

Access to Document

10.1117/12.2292188

https://arxiv.org/pdf/1803.07991

Cite this

Quantification of Lung Abnormalities in Cystic Fibrosis using Deep Networks. / Marques, Filipe; Dubost, Florian; Corput, Mariette Kemner-van de et al.
Medical Imaging 2018: Image Processing. SPIE - International Society for Optical Engineering, 2018. 105741G (Proceedings of SPIE International Symposium on Medical Imaging, Vol. 10574).

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Marques, F, Dubost, F, Corput, MKD, Tiddens, HAW & Bruijne, MD 2018, Quantification of Lung Abnormalities in Cystic Fibrosis using Deep Networks. in Medical Imaging 2018: Image Processing., 105741G, SPIE - International Society for Optical Engineering, Proceedings of SPIE International Symposium on Medical Imaging, vol. 10574, SPIE Medical Imaging 2018, Houston, United States, 10/02/2018. https://doi.org/10.1117/12.2292188

@inproceedings{c53a6edf8483445a87b1d16f3c711fe9,

title = "Quantification of Lung Abnormalities in Cystic Fibrosis using Deep Networks",

abstract = "Cystic fibrosis is a genetic disease which may appear in early life with structural abnormalities in lung tissues. We propose to detect these abnormalities using a texture classification approach. Our method is a cascade of two convolutional neural networks. The first network detects the presence of abnormal tissues. The second network identifies the type of the structural abnormalities: bronchiectasis, atelectasis or mucus plugging.We also propose a network computing pixel-wise heatmaps of abnormality presence learning only from the patch-wise annotations. Our database consists of CT scans of 194 subjects. We use 154 subjects to train our algorithms and the 40 remaining ones as a test set. We compare our method with random forest and a single neural network approach. The first network reaches a sensitivity of 0,62 for disease detection, 0,10 higher than the random forest classifier and 0,17 higher than the single neural network. Our cascade approach yields a final class-averaged F1-score of 0,38, outperforming the baseline method and the single network by 0,15 and 0,10.",

keywords = "cs.CV",

author = "Filipe Marques and Florian Dubost and Corput, {Mariette Kemner-van de} and Tiddens, {Harm A. W.} and Bruijne, {Marleen de}",

note = "SPIE - Medical Imaging 2018: Image Processing; SPIE Medical Imaging 2018 ; Conference date: 10-02-2018 Through 15-02-2018",

year = "2018",

doi = "10.1117/12.2292188",

language = "English",

series = "Proceedings of SPIE International Symposium on Medical Imaging",

booktitle = "Medical Imaging 2018",

publisher = "SPIE - International Society for Optical Engineering",

}

TY - GEN

T1 - Quantification of Lung Abnormalities in Cystic Fibrosis using Deep Networks

AU - Marques, Filipe

AU - Dubost, Florian

AU - Corput, Mariette Kemner-van de

AU - Tiddens, Harm A. W.

AU - Bruijne, Marleen de

N1 - SPIE - Medical Imaging 2018: Image Processing

PY - 2018

Y1 - 2018

N2 - Cystic fibrosis is a genetic disease which may appear in early life with structural abnormalities in lung tissues. We propose to detect these abnormalities using a texture classification approach. Our method is a cascade of two convolutional neural networks. The first network detects the presence of abnormal tissues. The second network identifies the type of the structural abnormalities: bronchiectasis, atelectasis or mucus plugging.We also propose a network computing pixel-wise heatmaps of abnormality presence learning only from the patch-wise annotations. Our database consists of CT scans of 194 subjects. We use 154 subjects to train our algorithms and the 40 remaining ones as a test set. We compare our method with random forest and a single neural network approach. The first network reaches a sensitivity of 0,62 for disease detection, 0,10 higher than the random forest classifier and 0,17 higher than the single neural network. Our cascade approach yields a final class-averaged F1-score of 0,38, outperforming the baseline method and the single network by 0,15 and 0,10.

AB - Cystic fibrosis is a genetic disease which may appear in early life with structural abnormalities in lung tissues. We propose to detect these abnormalities using a texture classification approach. Our method is a cascade of two convolutional neural networks. The first network detects the presence of abnormal tissues. The second network identifies the type of the structural abnormalities: bronchiectasis, atelectasis or mucus plugging.We also propose a network computing pixel-wise heatmaps of abnormality presence learning only from the patch-wise annotations. Our database consists of CT scans of 194 subjects. We use 154 subjects to train our algorithms and the 40 remaining ones as a test set. We compare our method with random forest and a single neural network approach. The first network reaches a sensitivity of 0,62 for disease detection, 0,10 higher than the random forest classifier and 0,17 higher than the single neural network. Our cascade approach yields a final class-averaged F1-score of 0,38, outperforming the baseline method and the single network by 0,15 and 0,10.

KW - cs.CV

U2 - 10.1117/12.2292188

DO - 10.1117/12.2292188

M3 - Article in proceedings

T3 - Proceedings of SPIE International Symposium on Medical Imaging

BT - Medical Imaging 2018

PB - SPIE - International Society for Optical Engineering

T2 - SPIE Medical Imaging 2018

Y2 - 10 February 2018 through 15 February 2018

ER -

Quantification of Lung Abnormalities in Cystic Fibrosis using Deep Networks

Abstract

Conference

Access to Document

Fingerprint

Cite this