Table 4. Diagnostic performances of different classification models through microcalcifications and mass features in combination (41 features).

	Test Dataset				Training Dataset
	accuracy	sensitivity	specificity	AUC	mean ± std (Accuracy)
SVM	85.8%	0.95	0.78	0.85	0.79 ± 0.07
KNN (N = 6)	84.3%	0.94	0.76	0.83	0.77 ± 0.06
LDA	74.0%	0.84	0.65	0.74	0.69 ± 0.07
SAE	89.7%	0.89	0.90	0.90	0.85 ± 0.06

The proposed SAE achieved superior performance in terms of the four measurements. The best measurements were highlighted in bold.