Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning

Gustavo Ulloa; Alejandro Veloz; Héctor Allende-Cid; Héctor Allende

doi:10.1007/978-3-030-50516-5_16

Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning

Gustavo Ulloa, Alejandro Veloz, Héctor Allende-Cid, Héctor Allende

SCHOOL OF COMPUTER ENGINEERING

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

5 Scopus citations

Abstract

Multiple sclerosis lesions segmentation is an important step in the diagnosis and tracking in the evolution of the disease. Convolutional Neural Networks (CNN) have been obtaining successful results in the task of lesion segmentation in recent years, but still present problem segmenting boundaries of the lesions. In this work we focus the learning process on hard voxels close to the boundaries of the lesions by means of a stratified sampling and the use of focal loss function that dynamically increase the penalization on this kind of voxels. This approach was applied on the 2015 Longitudinal MS Lesion Segmentation Challenge dataset (ISBI2015 (https://smart-stats-tools.org/lesion-challenge)), obtaining better results than approaches using binary cross entropy loss and focal loss functions with uniform sampling.

Original language	English
Title of host publication	Image Analysis and Recognition - 17th International Conference, ICIAR 2020, Proceedings
Editors	Aurélio Campilho, Fakhri Karray, Zhou Wang
Publisher	Springer
Pages	182-192
Number of pages	11
ISBN (Print)	9783030505158
DOIs	https://doi.org/10.1007/978-3-030-50516-5_16
State	Published - 2020
Event	17th International Conference on Image Analysis and Recognition, ICIAR 2020 - Póvoa de Varzim, Portugal Duration: 24 Jun 2020 → 26 Jun 2020

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12132 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	17th International Conference on Image Analysis and Recognition, ICIAR 2020
Country/Territory	Portugal
City	Póvoa de Varzim
Period	24/06/20 → 26/06/20

Keywords

Convolutional Neural Networks
Focal loss
Image segmentation
Magnetic Resonance Imaging
Multiple sclerosis lesions
Stratified sampling

Access to Document

10.1007/978-3-030-50516-5_16

Cite this

Ulloa, G., Veloz, A., Allende-Cid, H., & Allende, H. (2020). Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning. In A. Campilho, F. Karray, & Z. Wang (Eds.), Image Analysis and Recognition - 17th International Conference, ICIAR 2020, Proceedings (pp. 182-192). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12132 LNCS). Springer. https://doi.org/10.1007/978-3-030-50516-5_16

Ulloa, Gustavo ; Veloz, Alejandro ; Allende-Cid, Héctor et al. / Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning. Image Analysis and Recognition - 17th International Conference, ICIAR 2020, Proceedings. editor / Aurélio Campilho ; Fakhri Karray ; Zhou Wang. Springer, 2020. pp. 182-192 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{143a63a2c67b4dc2a063dba034b07712,

title = "Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning",

abstract = "Multiple sclerosis lesions segmentation is an important step in the diagnosis and tracking in the evolution of the disease. Convolutional Neural Networks (CNN) have been obtaining successful results in the task of lesion segmentation in recent years, but still present problem segmenting boundaries of the lesions. In this work we focus the learning process on hard voxels close to the boundaries of the lesions by means of a stratified sampling and the use of focal loss function that dynamically increase the penalization on this kind of voxels. This approach was applied on the 2015 Longitudinal MS Lesion Segmentation Challenge dataset (ISBI2015 (https://smart-stats-tools.org/lesion-challenge)), obtaining better results than approaches using binary cross entropy loss and focal loss functions with uniform sampling.",

keywords = "Convolutional Neural Networks, Focal loss, Image segmentation, Magnetic Resonance Imaging, Multiple sclerosis lesions, Stratified sampling",

author = "Gustavo Ulloa and Alejandro Veloz and H{\'e}ctor Allende-Cid and H{\'e}ctor Allende",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2020. Copyright: Copyright 2020 Elsevier B.V., All rights reserved.; 17th International Conference on Image Analysis and Recognition, ICIAR 2020 ; Conference date: 24-06-2020 Through 26-06-2020",

year = "2020",

doi = "10.1007/978-3-030-50516-5_16",

language = "English",

isbn = "9783030505158",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "182--192",

editor = "Aur{\'e}lio Campilho and Fakhri Karray and Zhou Wang",

booktitle = "Image Analysis and Recognition - 17th International Conference, ICIAR 2020, Proceedings",

}

Ulloa, G, Veloz, A, Allende-Cid, H & Allende, H 2020, Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning. in A Campilho, F Karray & Z Wang (eds), Image Analysis and Recognition - 17th International Conference, ICIAR 2020, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12132 LNCS, Springer, pp. 182-192, 17th International Conference on Image Analysis and Recognition, ICIAR 2020, Póvoa de Varzim, Portugal, 24/06/20. https://doi.org/10.1007/978-3-030-50516-5_16

Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning. / Ulloa, Gustavo; Veloz, Alejandro; Allende-Cid, Héctor et al.
Image Analysis and Recognition - 17th International Conference, ICIAR 2020, Proceedings. ed. / Aurélio Campilho; Fakhri Karray; Zhou Wang. Springer, 2020. p. 182-192 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12132 LNCS).

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

TY - GEN

T1 - Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning

AU - Ulloa, Gustavo

AU - Veloz, Alejandro

AU - Allende-Cid, Héctor

AU - Allende, Héctor

PY - 2020

Y1 - 2020

N2 - Multiple sclerosis lesions segmentation is an important step in the diagnosis and tracking in the evolution of the disease. Convolutional Neural Networks (CNN) have been obtaining successful results in the task of lesion segmentation in recent years, but still present problem segmenting boundaries of the lesions. In this work we focus the learning process on hard voxels close to the boundaries of the lesions by means of a stratified sampling and the use of focal loss function that dynamically increase the penalization on this kind of voxels. This approach was applied on the 2015 Longitudinal MS Lesion Segmentation Challenge dataset (ISBI2015 (https://smart-stats-tools.org/lesion-challenge)), obtaining better results than approaches using binary cross entropy loss and focal loss functions with uniform sampling.

AB - Multiple sclerosis lesions segmentation is an important step in the diagnosis and tracking in the evolution of the disease. Convolutional Neural Networks (CNN) have been obtaining successful results in the task of lesion segmentation in recent years, but still present problem segmenting boundaries of the lesions. In this work we focus the learning process on hard voxels close to the boundaries of the lesions by means of a stratified sampling and the use of focal loss function that dynamically increase the penalization on this kind of voxels. This approach was applied on the 2015 Longitudinal MS Lesion Segmentation Challenge dataset (ISBI2015 (https://smart-stats-tools.org/lesion-challenge)), obtaining better results than approaches using binary cross entropy loss and focal loss functions with uniform sampling.

KW - Convolutional Neural Networks

KW - Focal loss

KW - Image segmentation

KW - Magnetic Resonance Imaging

KW - Multiple sclerosis lesions

KW - Stratified sampling

UR - http://www.scopus.com/inward/record.url?scp=85087281353&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-50516-5_16

DO - 10.1007/978-3-030-50516-5_16

M3 - Conference contribution

AN - SCOPUS:85087281353

SN - 9783030505158

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

SP - 182

EP - 192

BT - Image Analysis and Recognition - 17th International Conference, ICIAR 2020, Proceedings

A2 - Campilho, Aurélio

A2 - Karray, Fakhri

A2 - Wang, Zhou

PB - Springer

T2 - 17th International Conference on Image Analysis and Recognition, ICIAR 2020

Y2 - 24 June 2020 through 26 June 2020

ER -

Ulloa G, Veloz A, Allende-Cid H, Allende H. Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning. In Campilho A, Karray F, Wang Z, editors, Image Analysis and Recognition - 17th International Conference, ICIAR 2020, Proceedings. Springer. 2020. p. 182-192. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-50516-5_16

Improving multiple sclerosis lesion boundaries segmentation by convolutional neural networks with focal learning

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this