Automatic segmentation and classification of breast lesions through identification of informative multiparametric PET/MRI features

European Radiology Experimental

Table 6 The ten top-ranked classification features according to Gini importance and minimum-redundancy-maximum-relevance

Rank	Manual annotation		Automatic segmentation
Rank	Gini importance	Minimum-redundancy-maximum-relevance	Gini importance	Minimum-redundancy-maximum-relevance
1	f_{l-tex-post/peak} entropy	f_{l-tex-post/peak} energy	f _l-kinetic T _1/2	f_{l-tex-post/peak} energy
2	f_{l-tex-post/peak} energy	f_l-δckc # 11	f_{l-tex-post/peak} energy	f_l-δckc #10
3	f _l-kinetic T _1/2	f_l-δckc # 8	f_l-kinetic MDER	f_{l-tex-post/pre} sum average
4	f_l-ckc #11	f_l-kinetic MDER	f _l-adc	f_{l-tex-post/pre} homogeneity
5	f_l-ckc #9	f_{l-tex-post/pre} homogeneity	f_l-δckc #18	f_l-δckc #20
6	f_l-tex-post information measure 1	f_{l-tex-post/peak} correlation	f_l-ckc #11	f_{l-tex-post/peak} information measure 2
7	f_l-δckc #14	f_{l-tex-post/peak} entropy	f_l-δckc #21	f_l-δckc #21
8	f_l-tex-post information measure 2	f _l-kinetic T _1/2	f_l-δckc #14	f _l-kinetic β
9	f_l-δckc #11	f_{l-tex-post/pre} dva	f_{l-tex-post/pre} dva	f_{l-tex-peak/pre} homogeneity
10	f_l-tex-peak entropy	f_{l-tex-post/pre} energy	f_l-δckc #10	f_l-tex-post sum variance

dva difference variance, MDER maximum derivative of kinetic regression function