Table 2.

Performance metrics for visual reading and the three machine learning classifiers.

A	Entire dataset
	AD				DLB				FTD				NC				Overall acc.
N	63				79				23				41
	F1	Sp	Pr	Se	F1	Sp	Pr	Se	F1	Sp	Pr	Se	F1	Sp	Pr	Se
Visual read	0.83	94	86	79	0.79	97	93	68	0.82	100	100	70	0.98	99	95	100	78
Pattern-based classifier	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
95 ROIs-based classifier	0.83	92	82	84	0.86	94	89	82	0.91	99	95	87	0.89	95	83	95	86
171 ROIs-based classifier	0.81	92	81	82	0.86	93	88	85	0.91	99	95	87	0.89	96	84	93	86
B	Reduced dataset
	AD				DLB				FTD				NC				Overall acc.
N	43				59				13				12
	F1	Sp	Pr	Se	F1	Sp	Pr	Se	F1	Sp	Pr	Se	F1	Sp	Pr	Se
Visual read	0.78	92	82	74	0.78	94	91	68	0.63	100	100	46	0.92	98	86	100	71
Pattern-based classifier	0.74	87	74	74	0.81	87	84	78	0.87	100	100	77	0.71	93	58	82	78
95 ROIs-based classifier	0.78	88	77	79	0.81	90	86	77	0.83	99	91	77	0.75	93	60	100	80
171 ROIs-based classifier	0.78	89	79	81	0.84	91	89	80	0.92	99	92	92	0.77	84	63	100	83

Performance metrics are calculated on (A) the entire dataset and on (B) the reduced dataset, where we excluded patients used for pattern identification from the testing set. Specificity, precision, sensitivity, and overall accuracy are presented as %. AD, dementia due to Alzheimer’s disease; DLB, dementia with Lewy bodies; FTD, frontotemporal dementia; NC, normal control; ROI, a region of interest; Sp, specificity; Pr, precision; Se, sensitivity; Acc, accuracy.