Table 4.

Multivariate models for method^a,b

	Features	Coefficient	P value	VIF
First-order model	90 keV mean	0.029	0.006	1.207
R 2 = 0.55	Iodine–calcium min	−0.017	0.04	1.080
R 2 (adjusted) = 0.39	BMI	0.019	0.005	1.144
Second-order model	90 keV mean	0.024	0.008	1.269
R 2 = 0.68	Iodine–calcium min	−0.025	0.01	1.944
R 2 (adjusted) = 0.57	Iodine–calcium max	0.024	0.01	2.912
	Iodine–water SD	−0.116	0.047	4.812
	Calcium–iodine max ^2	0.00185	0.01	1.582
	90 keV mean ^2	0.0016243	0.01	1.079
Second-order model with interactions	90 keV mean	0.0263	0.001	1.38
R 2 = 0.82	Iodine–calcium min	−0.0142	0.02	1.30
R 2 (adjusted) = 0.75	Iodine–calcium max	0.0107	0.04	1.63
	Weight	0.0100	0.047	1.70
	Calcium–iodine max × weight	0.00162	0.001	1.38
	90 keV mean × iodine–calcium min	0.00147	0.02	1.20
	Iodine–calcium max ^2	−0.000605	0.02	1.51

The significant value p <0.05

BMI body mass index, max maximum, min minimum, VIF variance inflation factor

^aPairs of substances represent the use of proprietary software designed to separate a pair of substances. For example, the iodine–calcium min is the lowest value pixel (minimum) on an image designed to separate iodine from calcium. On such an image, iodine will have higher values and calcium will have lower values. Conversely, on the calcium–iodine images, calcium will have higher values and iodine will have lower values. “Effective Z” refers to an image where the value is an estimate of the effective atomic number (Z); on these images, iodine will be higher than calcium, which will, in turn, be higher than the hydrogen, oxygen, nitrogen, and carbon that make up most of human tissue

^bVariables in italics are used in all three regression models