Table 5.

Classification-based accuracy comparisons between experimental sets. The scheme for calculating the accuracy for each classifier was the same as that indicated in Table 2

Study	Classification accuracy (%)				Improvement (%)	Accuracy degradation rate (%)
	SVM	p value	CNN	p value		SVM	CNN	p value
intrascanner set	89.53 ± 2.15	< 0.001	96.09 ± 1.41	< 0.001	6.56	1.61 ± 6.04	0.99 ± 5.40	0.03
integrated scanner set	88.07 ± 1.63		95.12 ± 1.91		7.05
		< 0.001		< 0.001		12.37 ± 9.51	9.98 ± 6.76	< 0.001
interscanner set	77.14 ± 3.25		85.59 ± 2.17		8.45