Table 2.

Classification performance for the sMCI vs. pMCI task: Accuracy (%), Sensitivity (%), and Specificity (%) of the proposed network compared with the published state-of-the-art methods for the task of discriminating between sMCI and pMCI subjects.

Method	Modality	Year to conversion	#Subjects	Accuracy	Sensitivity	Specificity
Young et al.18	MRI	0–3	143	64.3	53.2	69.8
Liu et al.40	MRI	0–3	234	68.8	64.29	74.07
Suk et al.28	MRI	unknown	204	72.42	36.7	90.98
Cheng et al.41	MRI	0–2	99	73.4	74.3	72.1
Zhu et al.42	MRI	0–1.5	99	71.8	48.0	92.8
Huang et al.46	longitudinal MRI	0–3	131	79.4	86.5	78.2
Proposed	MRI	0–3	626	75.44 (7.74)	73.27 (7.58)	76.19 (8.35)
Young et al.18	PET	0–3	143	65.0	66.0	64.6
Liu et al.40	PET	0–3	234	68.8	57.14	82.41
Suk et al.28	PET	unknown	204	70.75	25.45	96.55
Cheng et al.41	PET	0–2	99	71.6	76.4	67.9
Zhu et al.42	PET	0–1.5	99	71.2	47.4	93.0
Proposed	PET	0–3	626	81.53 (7.42)	78.20 (7.72)	82.47 (9.30)
Young et al.18	MRI + PET + APOE	0–3	143	69.9	78.7	65.6
Liu et al.40	PET + MRI	0–3	234	73.5	76.19	70.37
Suk et al.28	PET + MRI	unknown	204	75.92	48.04	95.23
Cheng et al.41	PET + MRI + CSF	0–2	99	79.4	84.5	72.7
Zhu et al.42	MRI + PET	0–1.5	99	72.4	49.1	94.6
Moradi et al.20	MRI + Age + cognitive measure	0–3	264	82	87	74
Xu et al.43	MRI + PET + florbetapir PET	0–3	110	77.8	74.1	81.5
Zhang et al.44	longitudinal MRI + PET	0–2	88	78.4	79.0	78.0
An et al.45	MRI + SNP	0–2	362	80.8	71.5	85.4
Korolev et al.21	MRI + Plasma + cognitive measure	0–3	259	80.0	83.0	76.0
Proposed	PET + MRI	0–3	626	82.93 (7.25)	79.69 (8.37)	83.84 (6.37)

‘PET’ in this table represents FDG-PET neuroimaging. Our proposed approach using deep neural networks was performed using a single FDG-PET image and a single T1-MRI acquired from each of 409 sMCI subjects and 217 pMCI subjects (total 626 subjects) and the average (standard-deviation) of the accuracy, sensitivity and specificity of classifier performance are reported.