TY - GEN
T1 - Spoke-LBP and ring-LBP
T2 - 12th IEEE International Symposium on Biomedical Imaging, ISBI 2015
AU - Wan, Sunhua
AU - Huang, Xiaolei
AU - Lee, Hsiang Chieh
AU - Fujimoto, James G.
AU - Zhou, Chao
N1 - Publisher Copyright:
© 2015 IEEE.
PY - 2015/7/21
Y1 - 2015/7/21
N2 - This paper proposes a texture feature which is applied on human breast Optical Coherence Microscopy (OCM) images to classify different types of breast tissues. Inspired by local binary pattern (LBP) texture features, a new variant of LBP feature, block based LBP (BLBP) is proposed. Instead of representing intensity differences between neighbors and a center pixel, BLBP feature extracts the intensity differences among certain blocks of the neighborhood around a pixel. Two different ways are proposed to organize the blocks: the spokes and the rings. By integrating spoke BLBP with ring BLBP features, very high classification accuracy is achieved using a neural network classifier. In one of our experiments which classifies 4310 OCM images into five tissue types, the classification accuracy increased from 81.7% to 92.4% when new features are used instead of the traditional LBP feature. In another experiment which classifies 46 large field OCM images as either benign or containing tumor, a classification accuracy of 91.3% is reached by using multi-scale BLBP features.
AB - This paper proposes a texture feature which is applied on human breast Optical Coherence Microscopy (OCM) images to classify different types of breast tissues. Inspired by local binary pattern (LBP) texture features, a new variant of LBP feature, block based LBP (BLBP) is proposed. Instead of representing intensity differences between neighbors and a center pixel, BLBP feature extracts the intensity differences among certain blocks of the neighborhood around a pixel. Two different ways are proposed to organize the blocks: the spokes and the rings. By integrating spoke BLBP with ring BLBP features, very high classification accuracy is achieved using a neural network classifier. In one of our experiments which classifies 4310 OCM images into five tissue types, the classification accuracy increased from 81.7% to 92.4% when new features are used instead of the traditional LBP feature. In another experiment which classifies 46 large field OCM images as either benign or containing tumor, a classification accuracy of 91.3% is reached by using multi-scale BLBP features.
KW - Local Binary Pattern (LBP)
KW - Optical Coherence Microscopy (OCM)
KW - Texture feature
KW - Tissue classification
KW - Tumor detection
UR - http://www.scopus.com/inward/record.url?scp=84944327170&partnerID=8YFLogxK
U2 - 10.1109/ISBI.2015.7163848
DO - 10.1109/ISBI.2015.7163848
M3 - Conference contribution
AN - SCOPUS:84944327170
T3 - Proceedings - International Symposium on Biomedical Imaging
SP - 195
EP - 199
BT - 2015 IEEE 12th International Symposium on Biomedical Imaging, ISBI 2015
PB - IEEE Computer Society
Y2 - 16 April 2015 through 19 April 2015
ER -