Segmenting a broad class of histological structures is a necessary to identify the presence of cancer, to clarify spatial relation between histological structures in the tumor environments, making precise medicine studies easy, and provide an exploratory tool for pathologists. Histological structure determination helps explain spatial tumor biology and adds an advantage for health care organizations. Role focuses on the segmentation of histological structures present in colored images with stains (H & E) of the breast tissue. Accurate segmentation of histological structures can help build a spatial interaction map identifying the relations between the pixels to serve as an exploratory tool for pathologists. Graph theory based methods proposed based on spatial color statistics and neighborhood of nuclei statistics as well as designed a new region-based score for evaluating segmentation algorithms. In the first method, pair wise pixel color statistics measures in an H&E optimized color space built to enhance the separation between hematoxylin and eosin stains. The first method is expected to be successful in segmenting structures with well-defined boundaries (e.g., adipose tissues, blood vessels).The second method is designed to segment large amorphous histological structures (e.g., tumor nests), the spatial statistics of inter-nuclei distances is considered. Working with expertly annotated breast H&E images, this paper demonstrated the ability of proposed algorithms to identify significant histological structures, and thus enable the understanding of their spatial relationships, and perhaps infer the status of the disease.

histopathalogical image analysis, image segmentation, image statistics

[1] Luong Nguyen, “Spatial statistics for segmenting histological structures in HE stained tissue images", IEEE TMI, 2017.
[2] E. Bejnordi et al, “Automated detection of dcis in whole-slide he stained breast histopathology images", IEEE TMI, 2016.
[3] J. Vicory et al, “Appearance normalization of histology slides, Computerized Medical Imaging and Graphics",vol. 43, pp. 8998, 2015.
[4] X. Li and K. N. Plataniotis, “A complete color normalization approach to histopathology images using color cues computed from saturation weighted statistics"IEEE TBME, vol. 62, no. 7, pp.18621873, 2015.
[5] P. Isola et al, “Crisp boundary detection using pointwise mutual information,"ECCV 2014, 2014, pp. 799814.
[6] Vahadane et al, “Structure-preserving color normalization and sparse stain separation for histological images,"IEEE TMI, 2016.
[7] F. Liu and L. Yang, “A novel cell detection method using deep convolutional neural network and maximum-weight independent set,"MICCAI, pp. 349357, 2015.
[8] J. L. Fine, “21st century workow: A proposal,"Journal of Pathology Informatics, vol. 5, no. 1,p. 44, 2014.
[9] B.-R. Wei and R. M. Simpson, “Digital pathology and image analysis augment biospecimen annotation and biobank quality assurance harmonization,",Clinical biochemistry, vol. 47, no. 4, pp.274279, 2014
[10] M. T. McCann et al, “Images as occlusions of textures: a framework for segmentation,", Clinical biochemistry, vol. 47, no. 4, pp. 274279, 2014.
[11] N. M. Rajpoot, “HyMaP: A hybrid magnitude-phase approach to unsupervised segmentation of tumor areas in breast cancer histology images,",Journal of Pathology Informatics, vol.4,no.2,p.1,2013