Improving Cancer Detection Classification Performance Using GANs in Breast Cancer Data

Authors: Strelcenia, E. and Prakoonwit, S.

Journal: IEEE Access

Volume: 11

Pages: 71594-71615

eISSN: 2169-3536

DOI: 10.1109/ACCESS.2023.3291336

Abstract:

Breast cancer is one of the most prevalent cancers in women. In recent years, many studies have been conducted in the breast cancer domain. Previous studies have confirmed that timely and accurate breast cancer detection allows patients to undergo early treatment. Recently, Generative Adversarial Networks have been applied in the medical domain to synthetically generate image and non-image data for diagnosis. However, the development of an effective classification model in healthcare is difficult owing to the limited datasets. To address this challenge, we propose a novel K-CGAN method trained in different settings to generate synthetic data. This study applied five classification methods and feature selection to non-image Wisconsin Breast Cancer data of 357 malignant and 212 benign cases for evaluation. Moreover, we used recall, precision, accuracy, and F1 Score on the synthetic data generated by the K-CGAN model to verify the classification performance of our proposed K-CGAN. The empirical study shows that K-CGAN performed well with the highest stability compared to the other GAN variants. Hence, our findings indicate that the synthetic data generated by K-CGAN accurately represent the original data.

https://eprints.bournemouth.ac.uk/38817/

Source: Scopus

Improving Cancer Detection Classification Performance Using GANs in Breast Cancer Data

Authors: Strelcenia, E. and Prakoonwit, S.

Journal: IEEE ACCESS

Volume: 11

Pages: 71594-71615

ISSN: 2169-3536

DOI: 10.1109/ACCESS.2023.3291336

https://eprints.bournemouth.ac.uk/38817/

Source: Web of Science (Lite)

Improving Cancer Detection Classification Performance Using GANs in Breast Cancer Data

Authors: Strelcenia, E. and Prakoonwit, S.

Journal: IEEE Access

Publisher: IEEE

eISSN: 2169-3536

ISSN: 2169-3536

DOI: 10.1109/ACCESS.2023.3291336

Abstract:

Breast cancer is one of the most prevalent cancers in women. In recent years, many studies have been conducted in the breast cancer domain. Previous studies have confirmed that timely and accurate breast cancer detection allows patients to undergo early treatment. Recently, Generative Adversarial Networks have been applied in the medical domain to synthetically generate image and non-image data for diagnosis. However, the development of an effective classification model in healthcare is difficult owing to the limited datasets. To address this challenge, we propose a novel K-CGAN method trained in different settings to generate synthetic data. This study applied five classification methods and feature selection to non-image Wisconsin Breast Cancer data of 357 malignant and 212 benign cases for evaluation. Moreover, we used recall, precision, accuracy, and F1 Score on the synthetic data generated by the K-CGAN model to verify the classification performance of our proposed K-CGAN. The empirical study shows that K-CGAN performed well with the highest stability compared to the other GAN variants. Hence, our findings indicate that the synthetic data generated by K-CGAN accurately represent the original data.

https://eprints.bournemouth.ac.uk/38817/

Source: Manual

Improving Cancer Detection Classification Performance Using GANs in Breast Cancer Data

Authors: Strelcenia, E. and Prakoonwit, S.

Journal: IEEE Access

Volume: 11

Pages: 71594-71615

Publisher: IEEE

ISSN: 2169-3536

Abstract:

Breast cancer is one of the most prevalent cancers in women. In recent years, many studies have been conducted in the breast cancer domain. Previous studies have confirmed that timely and accurate breast cancer detection allows patients to undergo early treatment. Recently, Generative Adversarial Networks have been applied in the medical domain to synthetically generate image and non-image data for diagnosis. However, the development of an effective classification model in healthcare is difficult owing to the limited datasets. To address this challenge, we propose a novel K-CGAN method trained in different settings to generate synthetic data. This study applied five classification methods and feature selection to non-image Wisconsin Breast Cancer data of 357 malignant and 212 benign cases for evaluation. Moreover, we used recall, precision, accuracy, and F1 Score on the synthetic data generated by the K-CGAN model to verify the classification performance of our proposed K-CGAN. The empirical study shows that K-CGAN performed well with the highest stability compared to the other GAN variants. Hence, our findings indicate that the synthetic data generated by K-CGAN accurately represent the original data.

https://eprints.bournemouth.ac.uk/38817/

Source: BURO EPrints