TABLE 3

Classification by Logistic Regression

Classification	PET parameter	Discriminating variable*	Odds ratio^†	P value^‡	Sensitivity^§ (%)	Specificity^§ (%)
FTD/AD/DLB vs. NC	K₁	BA40 K₁	0.81	<0.0001	90.4	84.2
		¹⁸F-FDG	BA40 ¹⁸F-FDG	0.72	0.0005	92.3	78.9
		K₁+DV^‖	BA40 K₁	0.81	0.0002	88.5	84.2
			DV	0.99	0.396
		¹⁸F-FDG+DV	BA40 ¹⁸F-FDG	0.72	0.0012	90.4	78.9
			DV	1.00	0.881
AD/DLB vs. FTD	K₁	BA24–BA31 K₁	1.13	0.0116	73.3	57.1
		¹⁸F-FDG	BA24–BA31 ¹⁸F-FDG	1.14	0.0087	77.8	71.4
		K₁+DV	BA24–BA31 K₁	1.14	0.0320	73.3	57.1
			DV	0.99	0.286
		¹⁸F-FDG+DV	BA24–BA31 ¹⁸F-FDG	1.12	0.0282	68.9	57.1
			DV	0.99	0.345
DLB vs. AD	K₁	BA24–BA17 K₁	1.12	0.0007	70.0	76.0
		¹⁸F-FDG	BA24–BA17 ¹⁸F-FDG	1.28	0.0020	85.0	84.0
		K₁+DV	BA24–BA17 K₁	1.15	0.311	90.0	96.0
			DV	0.84	0.178
		¹⁸F-FDG+DV	BA24–BA17 ¹⁸F-FDG	1.04	0.737	90.0	96.0
			DV	0.83	0.159

↵* BA(s) that provided best discrimination for each given test.
↵^† An increase of 0.01 in the discriminating variable multiplies the odds of being in the first of the 2 comparison groups by this factor. A value of 1.0 corresponds to lack of discrimination. The magnitude of deviation from 1.0 is a measure of the ability of variable to discriminate the 2 groups.
↵^‡ P value for a test of null hypothesis that odds ratio of the discriminating variable is 1.0.
↵^§ Sensitivity and specificity are based on cross-validation—that is, when each case is classified, logistic regression is recomputed with that case excluded.
↵^‖ DV is bilateral average of ¹¹C-DTBZ VMAT2 binding measures in putamen.