Fig 1.
a) Lymphocyte (LN), b) Normal Epithelial nuclei (EN), c) Cancerous Epithelial Nuclei (CN) and d) Mitotic nuclei (MN)
Fig 2.
Flowchart for the proposed nuclei segmentation technique.
Fig 3.
a) A Sub Image showing an overlapping nuclei cluster, b) pre-processing: enhanced grayscale image which is obtained by applying principal component analysis to image (a), c) Graphical illustration of stick tokens oriented perpendicular to gradient direction. d) Graphical illustration of stick tokens oriented parallel to gradient directions. e) Nuclei seed points plotted in red–found using non regional maximal suppression of the parallel saliency map f) result of parallel voting g) result of perpendicular voting, h) combined nuclei saliency map obtained by subtracting (g) from (f).
Fig 4.
a) Sub Image of a breast histopathology section b) Nuclei saliency map and c) Nuclei seed points
Fig 5.
a) Graphical illustration of boundary search paths. b) MRF formulation of the boundary delineation problem.
Fig 6.
Results of boundary detection.
Table 1.
Performance Analysis of Proposed Method.
Fig 7.
a) Original Image of grade 3 breast cancer histopathology sections, b) Corresponding segmentation results.
Fig 8.
a) WSI Image of a breast cancer histopathology slide, b) Segmentation result shown in yellow, c) Shown in green box is a 1000 x 1000 pixel patch selected starting at pixel position (12000, 15000)) and d) Closer view of the segmentation result on the region selected.
Fig 9.
Qualitative results (A)-(D) Original Images of Breast cancer Histopathology Images, (E)&(F) Segmentation result of Wienert et al. (2012), (G)&(H) Segmentation result of method Veta. et al. (2013), (I)-(L) Segmentation results of proposed method.
Table 2.
Performance evaluation and comparison of the proposed method on datasets from Wienert et al. (2012) and Veta. et al. (2013)