Cargando…

2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets

Deep learning (DL) models in breast ultrasound (BUS) image analysis face challenges with data imbalance and limited atypical tumor samples. Generative Adversarial Networks (GAN) address these challenges by providing efficient data augmentation for small datasets. However, current GAN approaches fail...

Descripción completa

Detalles Bibliográficos
Autores principales: Luo, Jie, Zhang, Heqing, Zhuang, Yan, Han, Lin, Chen, Ke, Hua, Zhan, Li, Cheng, Lin, Jiangli
Formato: Online Artículo Texto
Lenguaje:English
Publicado: MDPI 2023
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10610581/
https://www.ncbi.nlm.nih.gov/pubmed/37896706
http://dx.doi.org/10.3390/s23208614
_version_ 1785128290369929216
author Luo, Jie
Zhang, Heqing
Zhuang, Yan
Han, Lin
Chen, Ke
Hua, Zhan
Li, Cheng
Lin, Jiangli
author_facet Luo, Jie
Zhang, Heqing
Zhuang, Yan
Han, Lin
Chen, Ke
Hua, Zhan
Li, Cheng
Lin, Jiangli
author_sort Luo, Jie
collection PubMed
description Deep learning (DL) models in breast ultrasound (BUS) image analysis face challenges with data imbalance and limited atypical tumor samples. Generative Adversarial Networks (GAN) address these challenges by providing efficient data augmentation for small datasets. However, current GAN approaches fail to capture the structural features of BUS and generated images lack structural legitimacy and are unrealistic. Furthermore, generated images require manual annotation for different downstream tasks before they can be used. Therefore, we propose a two-stage GAN framework, 2s-BUSGAN, for generating annotated BUS images. It consists of the Mask Generation Stage (MGS) and the Image Generation Stage (IGS), generating benign and malignant BUS images using corresponding tumor contours. Moreover, we employ a Feature-Matching Loss (FML) to enhance the quality of generated images and utilize a Differential Augmentation Module (DAM) to improve GAN performance on small datasets. We conduct experiments on two datasets, BUSI and Collected. Moreover, results indicate that the quality of generated images is improved compared with traditional GAN methods. Additionally, our generated images underwent evaluation by ultrasound experts, demonstrating the possibility of deceiving doctors. A comparative evaluation showed that our method also outperforms traditional GAN methods when applied to training segmentation and classification models. Our method achieved a classification accuracy of 69% and 85.7% on two datasets, respectively, which is about 3% and 2% higher than that of the traditional augmentation model. The segmentation model trained using the 2s-BUSGAN augmented datasets achieved DICE scores of 75% and 73% on the two datasets, respectively, which were higher than the traditional augmentation methods. Our research tackles imbalanced and limited BUS image data challenges. Our 2s-BUSGAN augmentation method holds potential for enhancing deep learning model performance in the field.
format Online
Article
Text
id pubmed-10610581
institution National Center for Biotechnology Information
language English
publishDate 2023
publisher MDPI
record_format MEDLINE/PubMed
spelling pubmed-106105812023-10-28 2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets Luo, Jie Zhang, Heqing Zhuang, Yan Han, Lin Chen, Ke Hua, Zhan Li, Cheng Lin, Jiangli Sensors (Basel) Article Deep learning (DL) models in breast ultrasound (BUS) image analysis face challenges with data imbalance and limited atypical tumor samples. Generative Adversarial Networks (GAN) address these challenges by providing efficient data augmentation for small datasets. However, current GAN approaches fail to capture the structural features of BUS and generated images lack structural legitimacy and are unrealistic. Furthermore, generated images require manual annotation for different downstream tasks before they can be used. Therefore, we propose a two-stage GAN framework, 2s-BUSGAN, for generating annotated BUS images. It consists of the Mask Generation Stage (MGS) and the Image Generation Stage (IGS), generating benign and malignant BUS images using corresponding tumor contours. Moreover, we employ a Feature-Matching Loss (FML) to enhance the quality of generated images and utilize a Differential Augmentation Module (DAM) to improve GAN performance on small datasets. We conduct experiments on two datasets, BUSI and Collected. Moreover, results indicate that the quality of generated images is improved compared with traditional GAN methods. Additionally, our generated images underwent evaluation by ultrasound experts, demonstrating the possibility of deceiving doctors. A comparative evaluation showed that our method also outperforms traditional GAN methods when applied to training segmentation and classification models. Our method achieved a classification accuracy of 69% and 85.7% on two datasets, respectively, which is about 3% and 2% higher than that of the traditional augmentation model. The segmentation model trained using the 2s-BUSGAN augmented datasets achieved DICE scores of 75% and 73% on the two datasets, respectively, which were higher than the traditional augmentation methods. Our research tackles imbalanced and limited BUS image data challenges. Our 2s-BUSGAN augmentation method holds potential for enhancing deep learning model performance in the field. MDPI 2023-10-20 /pmc/articles/PMC10610581/ /pubmed/37896706 http://dx.doi.org/10.3390/s23208614 Text en © 2023 by the authors. https://creativecommons.org/licenses/by/4.0/Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
spellingShingle Article
Luo, Jie
Zhang, Heqing
Zhuang, Yan
Han, Lin
Chen, Ke
Hua, Zhan
Li, Cheng
Lin, Jiangli
2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets
title 2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets
title_full 2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets
title_fullStr 2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets
title_full_unstemmed 2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets
title_short 2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets
title_sort 2s-busgan: a novel generative adversarial network for realistic breast ultrasound image with corresponding tumor contour based on small datasets
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10610581/
https://www.ncbi.nlm.nih.gov/pubmed/37896706
http://dx.doi.org/10.3390/s23208614
work_keys_str_mv AT luojie 2sbusgananovelgenerativeadversarialnetworkforrealisticbreastultrasoundimagewithcorrespondingtumorcontourbasedonsmalldatasets
AT zhangheqing 2sbusgananovelgenerativeadversarialnetworkforrealisticbreastultrasoundimagewithcorrespondingtumorcontourbasedonsmalldatasets
AT zhuangyan 2sbusgananovelgenerativeadversarialnetworkforrealisticbreastultrasoundimagewithcorrespondingtumorcontourbasedonsmalldatasets
AT hanlin 2sbusgananovelgenerativeadversarialnetworkforrealisticbreastultrasoundimagewithcorrespondingtumorcontourbasedonsmalldatasets
AT chenke 2sbusgananovelgenerativeadversarialnetworkforrealisticbreastultrasoundimagewithcorrespondingtumorcontourbasedonsmalldatasets
AT huazhan 2sbusgananovelgenerativeadversarialnetworkforrealisticbreastultrasoundimagewithcorrespondingtumorcontourbasedonsmalldatasets
AT licheng 2sbusgananovelgenerativeadversarialnetworkforrealisticbreastultrasoundimagewithcorrespondingtumorcontourbasedonsmalldatasets
AT linjiangli 2sbusgananovelgenerativeadversarialnetworkforrealisticbreastultrasoundimagewithcorrespondingtumorcontourbasedonsmalldatasets