Table 3.

Ranking of each biomarker feature importance to prediction of diagnosis classification from the random forest analysis.

		CU vs. LMCI		LMCI vs. AD		CU vs. AD
	Rank	Biomarker feature	Relative importance (%)	Biomarker feature	Relative importance (%)	Biomarker feature	Relative importance (%)
Top half	1	Hippocampus volume	12.69 	FDG-temporal	 18.88	FDG-angular	 23.78
	2	Aβ-frontal	11.51	FDG-angular	 17.36	FDG-CingulumPost	 16.99
	3	Aβ-temporal	 8.57 	FDG-CingulumPost	 12.11	Hippocampus volume	 12.93
	4	FDG-angular	8.32	Hippocampus volume	 7.49	FDG-temporal	 10.00
	5	Entorhinal cortex volume	 7.88 	Aβ-precuneus	5.14	Aβ-temporal	 8.11
	6	Aβ-precuneus	7.78	Aβ-temporal	 4.97	Aβ-precuneus	 6.20
	7	pTau	7.60	pTau	 4.82	Entorhinal cortex volume	 4.81
	8	Aβ-cingulate	4.75	Entorhinal cortex volume	4.76	pTau	 3.92
	Subtotal		69.1		75.45		 86.74
Bottom half	9	Aβ-hippocampus	4.48	Aβ-parietal	4.71	Aβ-frontal	3.81
	10	Ventricles	4.35	Aβ-frontal	3.98	Aβ-parietal	 3.32
	11	FDG-CingulumPost	4.29	Ventricles	3.69	Aβ-Hippocampus	 3.17
	12	GMV	4.20	Aβ-Hippocampus	3.14	Aβ-Cingulate	 0.99
	13	WMV	3.56	Aβ-Cingulate	2.63	Ventricles	 0.65
	14	FDG-temporal	3.48	GMV	2.59	GMV	 0.60
	15	WBV	3.45	WBV	2.46	WMV	 0.38
	16	Aβ-parietal	3.10	WMV	1.29	WBV	 0.32
	Subtotal		30.9		24.55		13.26
Sum			100		100		100

CU, cognitively unimpaired; LMCI, late mild cognitive impairment; AD, Alzheimer’s disease; FDG, fluorodeoxyglucose; GMV, gray matter volume; WMV, white matter volume; WBV, whole brain volume.