Table 1.

Summary of extracted features. Features at regional level are computed at four different window sizes. Features from the full image represent average values.

Nuclei	Area	Eccentricity	Roundness	Centroid x, y
	Perimeter	Orientation	Major Axis	Minor Axis
	Mean Texture Contrast 1	Mean Texture Contrast 2	Mean Texture Homogenity 1	Mean Texture Homogenity 2
	Mean H value inside nuclei	Mean V value inside nuclei	Mean S value inside nuclei
Regional Concentrations	Stroma Ip	Background Iba	Nuclei Ib	Epithelial tissue from Ib
	Mean H in window	Mean S in window 2	Mean V in window
	Mean intensity Histogram H 1	Mean intensity Histogram H 2	Mean intensity Histogram H 3	Mean intensity Histogram H 4
	Mean intensity Histogram S 1	Mean intensity Histogram S 2	Mean intensity Histogram S 3	Mean intensity Histogram S 4
	Mean intensity Histogram V 1	Mean intensity Histogram V 2	Mean intensity Histogram V 3	Mean intensity Histogram V 4
Clusters (ducts)	Cluster area	Cluster roundness	Cells inside cluster	Distance to centroid
Global Image Concentrations	Stroma Ip	Background Iba	Nuclei Ib
	H value	V Value	S Value