Premium
Breast histopathology image segmentation using spatio‐colour‐texture based graph partition method
Author(s) -
BELSARE A.D.,
MUSHRIF M.M.,
PANGARKAR M.A.,
MESHRAM N.
Publication year - 2016
Publication title -
journal of microscopy
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.569
H-Index - 111
eISSN - 1365-2818
pISSN - 0022-2720
DOI - 10.1111/jmi.12361
Subject(s) - artificial intelligence , histopathology , segmentation , image segmentation , pattern recognition (psychology) , graph , partition (number theory) , image texture , computer science , computer vision , texture (cosmology) , graph partition , image (mathematics) , mathematics , combinatorics , pathology , medicine , theoretical computer science
Summary This paper proposes a novel integrated spatio‐colour‐texture based graph partitioning method for segmentation of nuclear arrangement in tubules with a lumen or in solid islands without a lumen from digitized Hematoxylin–Eosin stained breast histology images, in order to automate the process of histology breast image analysis to assist the pathologists. We propose a new similarity based super pixel generation method and integrate it with texton representation to form spatio‐colour‐texture map of Breast Histology Image. Then a new weighted distance based similarity measure is used for generation of graph and final segmentation using normalized cuts method is obtained. The extensive experiments carried shows that the proposed algorithm can segment nuclear arrangement in normal as well as malignant duct in breast histology tissue image. For evaluation of the proposed method the ground‐truth image database of 100 malignant and nonmalignant breast histology images is created with the help of two expert pathologists and the quantitative evaluation of proposed breast histology image segmentation has been performed. It shows that the proposed method outperforms over other methods.