Automatic Diagnosis of Alzheimer’s Disease and Mild Cognitive Impairment Based on CNN + SVM Networks with End-to-End Training

Table 8

Ablations studies of the proposed 3DCNN + SVM model applied to binary classification of MCI vs. NC (%).

Method	Training set				Testing set
	ACC	SEN	SPE	AUC	ACC	SEN	SPE	AUC
3DCNN	99.80	99.66	99.55	99.99	75.04	75.54	72.97	79.74
3DCNN + SVM	98.35	98.30	98.41	99.99	75.58	76.42	74.41	80.80
3DCNN + SVM + E2E	99.54	99.26	99.90	99.88	76.68	77.80	75.57	82.39