Polytomous diagnosis of ovarian tumors as benign, borderline, primary invasive or metastatic: development and validation of standard and kernel-based risk prediction models

Table 1 Overview of selected variables

		Dichotomous 1-versus-1 models
Variable	MLR	Ben vs Bord	Ben vs PrInv	Ben vs Meta	Bord vs PrInv	Bord vs Meta	PrInv vs Meta
Logistic regression-based variable selection
Ascites	×		×	×	×	×	×
Maximal diameter of solid part	×		×	×	×	×
Age	×		×	×
Entirely solid tumor	×			×		×	×
Irregular internal cyst walls	×		×	×
Personal history of ovarian cancer	×						×
Bilateral tumors	×				×
Maximal diameter of lesion	×	×
Papillary structures with blood flow	×	×	×
Unilocular tumor		×
R1U variable selection*
Ascites			×	×	×	×	×
Maximal diameter of solid part			×	×	×	×
Age		×	×		×
Entirely solid tumor				×		×	×
Irregular internal cyst walls			×	×	×
Personal history of ovarian cancer		×					×
Bilateral tumors					×
Maximal diameter of lesion		×
Papillary structures with blood flow		×
Number of papillations		×
Acoustic shadows			×

Ben: benign; PrInv: primary invasive; Bord: borderline; Meta: metastatic.
* R1U variable selection: variable selection within the framework of least squares support vector machines that is based on rank 1 updates of the kernel matrix [27].

ISSN: 1471-2288