Automatic Breast and Fibroglandular Tissue Segmentation in Breast MRI Using Deep Learning by a Fully-Convolutional Residual Neural Network U-Net

Yang Zhang, Jeon Hor Chen, Kai Ting Chang, Vivian Youngjean Park, minjung Kim, Siwa Chan, Peter Chang, Daniel Chow, Alex Luk, Tiffany Kwong, Min Ying Su

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Rationale and Objectives: Breast segmentation using the U-net architecture was implemented and tested in independent validation datasets to quantify fibroglandular tissue volume in breast MRI. Materials and Methods: Two datasets were used. The training set was MRI of 286 patients with unilateral breast cancer. The segmentation was done on the contralateral normal breasts. The ground truth for the breast and fibroglandular tissue (FGT) was obtained by using a template-based segmentation method. The U-net deep learning algorithm was implemented to analyze the training set, and the final model was obtained using 10-fold cross-validation. The independent validation set was MRI of 28 normal volunteers acquired using four different MR scanners. Dice Similarity Coefficient (DSC), voxel-based accuracy, and Pearson's correlation were used to evaluate the performance. Results: For the 10-fold cross-validation in the initial training set of 286 patients, the DSC range was 0.83–0.98 (mean 0.95 ± 0.02) for breast and 0.73–0.97 (mean 0.91 ± 0.03) for FGT; and the accuracy range was 0.92–0.99 (mean 0.98 ± 0.01) for breast and 0.87–0.99 (mean 0.97 ± 0.01) for FGT. For the entire 224 testing breasts of the 28 normal volunteers in the validation datasets, the mean DSC was 0.86 ± 0.05 for breast, 0.83 ± 0.06 for FGT; and the mean accuracy was 0.94 ± 0.03 for breast and 0.93 ± 0.04 for FGT. The testing results for MRI acquired using four different scanners were comparable. Conclusion: Deep learning based on the U-net algorithm can achieve accurate segmentation results for the breast and FGT on MRI. It may provide a reliable and efficient method to process large number of MR images for quantitative analysis of breast density.

Original languageEnglish
JournalAcademic Radiology
DOIs
Publication statusAccepted/In press - 2019 Jan 1

Fingerprint

Breast
Learning
Healthy Volunteers
Datasets

All Science Journal Classification (ASJC) codes

  • Radiology Nuclear Medicine and imaging

Cite this

Zhang, Yang ; Chen, Jeon Hor ; Chang, Kai Ting ; Park, Vivian Youngjean ; Kim, minjung ; Chan, Siwa ; Chang, Peter ; Chow, Daniel ; Luk, Alex ; Kwong, Tiffany ; Su, Min Ying. / Automatic Breast and Fibroglandular Tissue Segmentation in Breast MRI Using Deep Learning by a Fully-Convolutional Residual Neural Network U-Net. In: Academic Radiology. 2019.
@article{1ddfaf03cbf54815a306546c0dd25ba6,
title = "Automatic Breast and Fibroglandular Tissue Segmentation in Breast MRI Using Deep Learning by a Fully-Convolutional Residual Neural Network U-Net",
abstract = "Rationale and Objectives: Breast segmentation using the U-net architecture was implemented and tested in independent validation datasets to quantify fibroglandular tissue volume in breast MRI. Materials and Methods: Two datasets were used. The training set was MRI of 286 patients with unilateral breast cancer. The segmentation was done on the contralateral normal breasts. The ground truth for the breast and fibroglandular tissue (FGT) was obtained by using a template-based segmentation method. The U-net deep learning algorithm was implemented to analyze the training set, and the final model was obtained using 10-fold cross-validation. The independent validation set was MRI of 28 normal volunteers acquired using four different MR scanners. Dice Similarity Coefficient (DSC), voxel-based accuracy, and Pearson's correlation were used to evaluate the performance. Results: For the 10-fold cross-validation in the initial training set of 286 patients, the DSC range was 0.83–0.98 (mean 0.95 ± 0.02) for breast and 0.73–0.97 (mean 0.91 ± 0.03) for FGT; and the accuracy range was 0.92–0.99 (mean 0.98 ± 0.01) for breast and 0.87–0.99 (mean 0.97 ± 0.01) for FGT. For the entire 224 testing breasts of the 28 normal volunteers in the validation datasets, the mean DSC was 0.86 ± 0.05 for breast, 0.83 ± 0.06 for FGT; and the mean accuracy was 0.94 ± 0.03 for breast and 0.93 ± 0.04 for FGT. The testing results for MRI acquired using four different scanners were comparable. Conclusion: Deep learning based on the U-net algorithm can achieve accurate segmentation results for the breast and FGT on MRI. It may provide a reliable and efficient method to process large number of MR images for quantitative analysis of breast density.",
author = "Yang Zhang and Chen, {Jeon Hor} and Chang, {Kai Ting} and Park, {Vivian Youngjean} and minjung Kim and Siwa Chan and Peter Chang and Daniel Chow and Alex Luk and Tiffany Kwong and Su, {Min Ying}",
year = "2019",
month = "1",
day = "1",
doi = "10.1016/j.acra.2019.01.012",
language = "English",
journal = "Academic Radiology",
issn = "1076-6332",
publisher = "Elsevier USA",

}

Automatic Breast and Fibroglandular Tissue Segmentation in Breast MRI Using Deep Learning by a Fully-Convolutional Residual Neural Network U-Net. / Zhang, Yang; Chen, Jeon Hor; Chang, Kai Ting; Park, Vivian Youngjean; Kim, minjung; Chan, Siwa; Chang, Peter; Chow, Daniel; Luk, Alex; Kwong, Tiffany; Su, Min Ying.

In: Academic Radiology, 01.01.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Automatic Breast and Fibroglandular Tissue Segmentation in Breast MRI Using Deep Learning by a Fully-Convolutional Residual Neural Network U-Net

AU - Zhang, Yang

AU - Chen, Jeon Hor

AU - Chang, Kai Ting

AU - Park, Vivian Youngjean

AU - Kim, minjung

AU - Chan, Siwa

AU - Chang, Peter

AU - Chow, Daniel

AU - Luk, Alex

AU - Kwong, Tiffany

AU - Su, Min Ying

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Rationale and Objectives: Breast segmentation using the U-net architecture was implemented and tested in independent validation datasets to quantify fibroglandular tissue volume in breast MRI. Materials and Methods: Two datasets were used. The training set was MRI of 286 patients with unilateral breast cancer. The segmentation was done on the contralateral normal breasts. The ground truth for the breast and fibroglandular tissue (FGT) was obtained by using a template-based segmentation method. The U-net deep learning algorithm was implemented to analyze the training set, and the final model was obtained using 10-fold cross-validation. The independent validation set was MRI of 28 normal volunteers acquired using four different MR scanners. Dice Similarity Coefficient (DSC), voxel-based accuracy, and Pearson's correlation were used to evaluate the performance. Results: For the 10-fold cross-validation in the initial training set of 286 patients, the DSC range was 0.83–0.98 (mean 0.95 ± 0.02) for breast and 0.73–0.97 (mean 0.91 ± 0.03) for FGT; and the accuracy range was 0.92–0.99 (mean 0.98 ± 0.01) for breast and 0.87–0.99 (mean 0.97 ± 0.01) for FGT. For the entire 224 testing breasts of the 28 normal volunteers in the validation datasets, the mean DSC was 0.86 ± 0.05 for breast, 0.83 ± 0.06 for FGT; and the mean accuracy was 0.94 ± 0.03 for breast and 0.93 ± 0.04 for FGT. The testing results for MRI acquired using four different scanners were comparable. Conclusion: Deep learning based on the U-net algorithm can achieve accurate segmentation results for the breast and FGT on MRI. It may provide a reliable and efficient method to process large number of MR images for quantitative analysis of breast density.

AB - Rationale and Objectives: Breast segmentation using the U-net architecture was implemented and tested in independent validation datasets to quantify fibroglandular tissue volume in breast MRI. Materials and Methods: Two datasets were used. The training set was MRI of 286 patients with unilateral breast cancer. The segmentation was done on the contralateral normal breasts. The ground truth for the breast and fibroglandular tissue (FGT) was obtained by using a template-based segmentation method. The U-net deep learning algorithm was implemented to analyze the training set, and the final model was obtained using 10-fold cross-validation. The independent validation set was MRI of 28 normal volunteers acquired using four different MR scanners. Dice Similarity Coefficient (DSC), voxel-based accuracy, and Pearson's correlation were used to evaluate the performance. Results: For the 10-fold cross-validation in the initial training set of 286 patients, the DSC range was 0.83–0.98 (mean 0.95 ± 0.02) for breast and 0.73–0.97 (mean 0.91 ± 0.03) for FGT; and the accuracy range was 0.92–0.99 (mean 0.98 ± 0.01) for breast and 0.87–0.99 (mean 0.97 ± 0.01) for FGT. For the entire 224 testing breasts of the 28 normal volunteers in the validation datasets, the mean DSC was 0.86 ± 0.05 for breast, 0.83 ± 0.06 for FGT; and the mean accuracy was 0.94 ± 0.03 for breast and 0.93 ± 0.04 for FGT. The testing results for MRI acquired using four different scanners were comparable. Conclusion: Deep learning based on the U-net algorithm can achieve accurate segmentation results for the breast and FGT on MRI. It may provide a reliable and efficient method to process large number of MR images for quantitative analysis of breast density.

UR - http://www.scopus.com/inward/record.url?scp=85060764033&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85060764033&partnerID=8YFLogxK

U2 - 10.1016/j.acra.2019.01.012

DO - 10.1016/j.acra.2019.01.012

M3 - Article

JO - Academic Radiology

JF - Academic Radiology

SN - 1076-6332

ER -