Table 5.

All features whose selection frequency is significantly different from the chance in the benign/malignant classification (p-value null model test <0.05). Direction and level of the Haar decomposition, GLCM direction, and gradient’s magnitude or direction of the extracted features are shown.

Feature Set	Feature	ROI Type	Frequency (%)
COUNT	SIFT	LE	$84.8$
HAAR	Variance_LL2	RC	$50.4$
STAT	RelativeSmoothness	RC	$45.8$
GRAD	RelativeSmoothness_Gmag	LE	$43.8$
HAAR	Variance_LL1	RC	$34.8$
STAT	Variance	RC	$34.2$
GRAD	Variance_Gmag	LE	$31.9$
GLCM	ClusterProminence_HL1 ($\theta ={90}^{\circ }$)	RC	$31.0$
GLCM	Correlation_LH1 ($\theta =0^{\circ }$)	RC	$30.3$
HAAR	Variance_LH1	LE	$29.9$
HAAR	RelativeSmoothness_HL2	RC	$29.5$
HAAR	Variance_LH2	LE	$23.6$
GLCM	Homogeneity_LH1 ($\theta =0^{\circ }$)	RC	$22.4$
STAT	Standard Deviation	RC	$20.3$
HAAR	Variance_HL2	RC	$19.5$
STAT	Maximum − Minimum	RC	$17.8$
COUNT	MinEigen	RC	$16.4$