Table 3.

Top five weighted radiomic features of trained logistic regression classifiers. The feature names are based on the Pyradiomics library used for research [16].

Weight Ranking	Radiomic Feature Names
	L1 Penalty and F0		L1 Penalty and F4		L2 Penalty and F0		L2 Penalty and F4
	Feature Category	Feature Name	Feature Category	Feature Name	Feature Category	Feature Name	Feature Category	Feature Name
1st	GLRLM *	Low Gray Level Run Emphasis *	First order *	Robust Mean Absolute Deviation *	Shape *	Flatness *	NGTDM *	Coarseness *
2nd	Shape *	Flatness *	Shape #	Maximum 2D Diameter Slice #	Shape *	Least Axis Length *	Shape #	Maximum 2D Diameter Slice #
3rd	Shape *	Maximum 2D Diameter Row *	GLRLM *	Run Length Nonuniformity Normalized *	Shape *	Maximum 2D Diameter Row *	First order *	Robust Mean Absolute Deviation *
4th	GLSZ *	Zone Entropy *	NGTDM *	Coarseness *	GLRLM *	Low Gray Level Run Emphasis *	Shape #	Surface Area #
5th	Shape *	Least Axis Length *	GLCM #	Cluster Shade #	GLRLM *	Short Run Low Gray Level Emphasis *	Shape #	Minor Axis Length #

* liver radiomic features; ^# splenic radiomic features. Abbreviations: GLRLM = gray level run length matrix; NGTDM = neighboring gray tone difference matrix; GLSZM = gray level size zone matrix; GLCM = gray level co-occurrence matrix.