FCA-Net: Adversarial Learning for Skin Lesion Segmentation Based on Multi-Scale Features and Factorized Channel Attention

Skin lesion segmentation in dermoscopic images is still a challenge due to the low contrast and fuzzy boundaries of lesions. Moreover, lesions have high similarity with the healthy regions in terms of appearance. In this paper, we propose an accurate skin lesion segmentation model based on a modifie...

Full description

Bibliographic Details
Main Authors: Vivek Kumar Singh, Mohamed Abdel-Nasser, Hatem A. Rashwan, Farhan Akram, Nidhi Pandey, Alain Lalande, Benoit Presles, Santiago Romani, Domenec Puig
Format: Article
Language:English
Published: IEEE 2019-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/8832175/
_version_ 1818947311164719104
author Vivek Kumar Singh
Mohamed Abdel-Nasser
Hatem A. Rashwan
Farhan Akram
Nidhi Pandey
Alain Lalande
Benoit Presles
Santiago Romani
Domenec Puig
author_facet Vivek Kumar Singh
Mohamed Abdel-Nasser
Hatem A. Rashwan
Farhan Akram
Nidhi Pandey
Alain Lalande
Benoit Presles
Santiago Romani
Domenec Puig
author_sort Vivek Kumar Singh
collection DOAJ
description Skin lesion segmentation in dermoscopic images is still a challenge due to the low contrast and fuzzy boundaries of lesions. Moreover, lesions have high similarity with the healthy regions in terms of appearance. In this paper, we propose an accurate skin lesion segmentation model based on a modified conditional generative adversarial network (cGAN). We introduce a new block in the encoder of cGAN called factorized channel attention (FCA), which exploits both channel attention mechanism and residual 1-D kernel factorized convolution. The channel attention mechanism increases the discriminability between the lesion and non-lesion features by taking feature channel interdependencies into account. The 1-D factorized kernel block provides extra convolutions layers with a minimum number of parameters to reduce the computations of the higher-order convolutions. Besides, we use a multi-scale input strategy to encourage the development of filters which are scale-variant (i.e., constructing a scale-invariant representation). The proposed model is assessed on three skin challenge datasets: ISBI2016, ISBI2017, and ISIC2018. It yields competitive results when compared to several state-of-the-art methods in terms of Dice coefficient and intersection over union (IoU) score. The codes of the proposed model are publicly available at https://github.com/vivek231/Skin-Project.
first_indexed 2024-12-20T08:28:53Z
format Article
id doaj.art-a437eed0236b400f940e8694a32bb5da
institution Directory Open Access Journal
issn 2169-3536
language English
last_indexed 2024-12-20T08:28:53Z
publishDate 2019-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj.art-a437eed0236b400f940e8694a32bb5da2022-12-21T19:46:46ZengIEEEIEEE Access2169-35362019-01-01713055213056510.1109/ACCESS.2019.29404188832175FCA-Net: Adversarial Learning for Skin Lesion Segmentation Based on Multi-Scale Features and Factorized Channel AttentionVivek Kumar Singh0https://orcid.org/0000-0002-8259-7087Mohamed Abdel-Nasser1Hatem A. Rashwan2https://orcid.org/0000-0001-5421-1637Farhan Akram3Nidhi Pandey4Alain Lalande5Benoit Presles6Santiago Romani7Domenec Puig8Department of Computer Engineering and Mathematics, Rovira i Virgili University, Tarragona, SpainDepartment of Computer Engineering and Mathematics, Rovira i Virgili University, Tarragona, SpainDepartment of Computer Engineering and Mathematics, Rovira i Virgili University, Tarragona, SpainImaging Informatics Division, Bioinformatics Institute, A*STAR, Singapore138671Department of Computer Engineering and Mathematics, Rovira i Virgili University, Tarragona, SpainImViA EA 7535, University of Burgundy, Dijon, FranceImViA EA 7535, University of Burgundy, Dijon, FranceDepartment of Computer Engineering and Mathematics, Rovira i Virgili University, Tarragona, SpainDepartment of Computer Engineering and Mathematics, Rovira i Virgili University, Tarragona, SpainSkin lesion segmentation in dermoscopic images is still a challenge due to the low contrast and fuzzy boundaries of lesions. Moreover, lesions have high similarity with the healthy regions in terms of appearance. In this paper, we propose an accurate skin lesion segmentation model based on a modified conditional generative adversarial network (cGAN). We introduce a new block in the encoder of cGAN called factorized channel attention (FCA), which exploits both channel attention mechanism and residual 1-D kernel factorized convolution. The channel attention mechanism increases the discriminability between the lesion and non-lesion features by taking feature channel interdependencies into account. The 1-D factorized kernel block provides extra convolutions layers with a minimum number of parameters to reduce the computations of the higher-order convolutions. Besides, we use a multi-scale input strategy to encourage the development of filters which are scale-variant (i.e., constructing a scale-invariant representation). The proposed model is assessed on three skin challenge datasets: ISBI2016, ISBI2017, and ISIC2018. It yields competitive results when compared to several state-of-the-art methods in terms of Dice coefficient and intersection over union (IoU) score. The codes of the proposed model are publicly available at https://github.com/vivek231/Skin-Project.https://ieeexplore.ieee.org/document/8832175/Skin lesionconditional generative adversarial networkchannel attentionfactorized kernelresidual convolution
spellingShingle Vivek Kumar Singh
Mohamed Abdel-Nasser
Hatem A. Rashwan
Farhan Akram
Nidhi Pandey
Alain Lalande
Benoit Presles
Santiago Romani
Domenec Puig
FCA-Net: Adversarial Learning for Skin Lesion Segmentation Based on Multi-Scale Features and Factorized Channel Attention
IEEE Access
Skin lesion
conditional generative adversarial network
channel attention
factorized kernel
residual convolution
title FCA-Net: Adversarial Learning for Skin Lesion Segmentation Based on Multi-Scale Features and Factorized Channel Attention
title_full FCA-Net: Adversarial Learning for Skin Lesion Segmentation Based on Multi-Scale Features and Factorized Channel Attention
title_fullStr FCA-Net: Adversarial Learning for Skin Lesion Segmentation Based on Multi-Scale Features and Factorized Channel Attention
title_full_unstemmed FCA-Net: Adversarial Learning for Skin Lesion Segmentation Based on Multi-Scale Features and Factorized Channel Attention
title_short FCA-Net: Adversarial Learning for Skin Lesion Segmentation Based on Multi-Scale Features and Factorized Channel Attention
title_sort fca net adversarial learning for skin lesion segmentation based on multi scale features and factorized channel attention
topic Skin lesion
conditional generative adversarial network
channel attention
factorized kernel
residual convolution
url https://ieeexplore.ieee.org/document/8832175/
work_keys_str_mv AT vivekkumarsingh fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention
AT mohamedabdelnasser fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention
AT hatemarashwan fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention
AT farhanakram fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention
AT nidhipandey fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention
AT alainlalande fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention
AT benoitpresles fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention
AT santiagoromani fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention
AT domenecpuig fcanetadversariallearningforskinlesionsegmentationbasedonmultiscalefeaturesandfactorizedchannelattention