. 2017 Oct 17;31(4):415–424. doi: 10.1007/s10278-017-0028-9

Table 3.

Classification error rates based on the CNN between subregions with increasing numbers of convolution layers

Study set	No. of convolution layers	Classification accuracy (%)	Error rate (%)
Study set	No. of convolution layers	Classification accuracy (%)	N > E^a	E > N^b	H > R^c	R > H^d
Integrated set	1	81.27 ± 1.57	14.92	22.81	14.71	17.11
	2	90.67 ± 1.02	5.44	7.69	8.46	7.29
	3	93.73 ± 0.71	2.50	1.46	7.14	6.13
	4	95.12 ± 0.52	0.48	1.21	2.12	4.91

^aNormal case was misclassified as an emphysema case

^bEmphysema case was misclassified as a normal case

^cHoneycombing case was misclassified as a reticular opacity case

^dReticular opacity case was misclassified as a honeycombing case