TY - GEN
T1 - Medical Knowledge Constrained Semantic Breast Ultrasound Image Segmentation
AU - Huang, Kuan
AU - Cheng, H. D.
AU - Zhang, Yingtao
AU - Zhang, Boyu
AU - Xing, Ping
AU - Ning, Chunping
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/11/26
Y1 - 2018/11/26
N2 - Computer-aided diagnosis (CAD) can help doctors in diagnosing breast cancer. Breast ultrasound (BUS) imaging is harmless, effective, portable, and is the most popular modality for breast cancer detection/diagnosis. Many researchers work on improving the performance of CAD systems. However, there are two main shortcomings: (1) Most of the existing methods are based on prerequisites that there is one and only one tumor in the image. (2) The results depend on the datasets, i.e., an algorithm using different datasets may obtain different performances. It implies that the performance of traditional methods is dataset-dependent. In this paper, we propose an effective approach: (1) using information extended images to train a fully convolutional network (FCN) to semantically segment BUS image into 3 categories: Mammary layer, tumor, and background; and (2) applying layer structure information - the breast cancers are located inside the mammary layer - to the conditional random field (CRF) for conducting breast cancer segmentation and making the segmentation result more accurate. The proposed method is evaluated utilizing BUS images of 325 cases, and the result is the best comparing with that of the existing methods by achieving true positive rate 92.80%, false positive rate 9%, and Intersection over Union 82.11%. The proposed approach has solved the above mentioned two shortcomings of the existing methods.
AB - Computer-aided diagnosis (CAD) can help doctors in diagnosing breast cancer. Breast ultrasound (BUS) imaging is harmless, effective, portable, and is the most popular modality for breast cancer detection/diagnosis. Many researchers work on improving the performance of CAD systems. However, there are two main shortcomings: (1) Most of the existing methods are based on prerequisites that there is one and only one tumor in the image. (2) The results depend on the datasets, i.e., an algorithm using different datasets may obtain different performances. It implies that the performance of traditional methods is dataset-dependent. In this paper, we propose an effective approach: (1) using information extended images to train a fully convolutional network (FCN) to semantically segment BUS image into 3 categories: Mammary layer, tumor, and background; and (2) applying layer structure information - the breast cancers are located inside the mammary layer - to the conditional random field (CRF) for conducting breast cancer segmentation and making the segmentation result more accurate. The proposed method is evaluated utilizing BUS images of 325 cases, and the result is the best comparing with that of the existing methods by achieving true positive rate 92.80%, false positive rate 9%, and Intersection over Union 82.11%. The proposed approach has solved the above mentioned two shortcomings of the existing methods.
KW - breast ultrasound (BUS) image
KW - computer aided diagnosis (CAD)
KW - conditional random field (CRF)
KW - deep convolutional neural network (DCNN)
KW - semantic image segmentation
UR - http://www.scopus.com/inward/record.url?scp=85059760860&partnerID=8YFLogxK
U2 - 10.1109/ICPR.2018.8545272
DO - 10.1109/ICPR.2018.8545272
M3 - Conference contribution
AN - SCOPUS:85059760860
T3 - Proceedings - International Conference on Pattern Recognition
SP - 1193
EP - 1198
BT - 2018 24th International Conference on Pattern Recognition, ICPR 2018
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 24th International Conference on Pattern Recognition, ICPR 2018
Y2 - 20 August 2018 through 24 August 2018
ER -