Predicting Breast Density of Digital Breast Tomosynthesis from 2D Mammograms

Abstract

Breast density may be used as a predictor of breast cancer risk and can measure the condition of tissues on mammograms. This research developed a computer-aided diagnosis (CAD) system to predict breast density on digital breast tomosynthesis (DBT) images. We used two-dimensional (2D) mammograms to train the linear discriminant analysis (LDA) classifier. Then load the DBT projection image to predict breast density. Experimental results show that LDA is better than other classification methods, such as one rule, naive Bayes, decision tree and support vector machine. The breast density prediction accuracy of DBT is 80%.

KEYWORDS:

• Breast density
• Computer-aided diagnosis
• Classification
• Digital breast tomosynthesis
• Linear discrimination analysis
• Mammograms

Acknowledgements

We acknowledge the 2016 International Conference on Innovation and Management (IAM 2016 Summer) for publishing this abstract in their conference proceedings.

Additional information

Funding

This work was supported by Ministry of Science and Technology, Taiwan [grant number MOST 105-2221-E-415-009-MY2].

Notes on contributors

Jinn-Yi Yeh

Jinn-Yi Yeh received the BS degree from Chung-Yuan Christian University, Taiwan, in 1987 and the MS and PhD degrees from the University of Texas at Arlington in 1993 and 1997, respectively. He is presently a professor in the Department of Management Information Systems, National Chiayi University, Taiwan. His research interests focus on data mining, medical image processing, and deep learning. Email: [email protected]

Tu-Liang Lin

Tu-Liang Lin is an associate professor in the Department of Management Information Systems at National Chiayi University, Taiwan. He received the Ph.D. degree from the Iowa State University in 2011, under the supervision of Prof. Guang Song. His main research interests include robotics, wireless network, network security, bioinformatics and machine learning. He applied the robotic motion planning to discover the ligand migration path of dynamic proteins in his doctoral work. Currently, he is working on applying the deep learning techniques to solve the medical and biology problems. Corresponding author. Email: [email protected]

Siwa Chan

Siwa Chan received the Bachelor of Medicine from Kaohsiung Medical University, and MS of health and medical engineering from Feng Chia University, Taiwan. She is a doctor candidate in the Department of Electronic and Informatics, National Taiwan University. She is presently an MD in the Department of Medical Imaging, Taichung Tzu Chi Hospital, Taiwan. Email: [email protected]