Table 6.

Diagnostic performance comparison for both classification stages between the developed RC-CAD system and the state-of-the-art approaches by [27,32,37]. The diagnostic abilities of the RC-CAD outperformed all other methods in both classification stages. Let Sens: sensitivity, Spec: Specificity, DSC: Dice similarity coefficient, and Acc: Accuracy.

First Stage Classification (RCC vs. AML) Performance (Mean ± SD ≈)
Method		Sens%	Spec%	DSC
RC-CAD (Proposed)		95.3 ± 2.0	99.9 ± 0.4	0.98 ± 0.01
Kunapuli [27]		81.4 ± 0.0	95.7 ± 0.0	0.88 ± 0.00
Oberai [37]		88.9 ± 1.7	87.4 ± 1.4	0.91 ± 0.01
Lee [32]	AlexNet	84.0 ± 1.7	93.4 ± 1.9	0.88 ± 0.02
	GoogleNet	88.3 ± 1.7	95.1 ± 1.9	0.91 ± 0.01
	ResNet	88.0 ± 3.5	95.7 ± 0.9	0.91 ± 0.02
	VGGNet	86.9 ± 0.6	91.4 ± 2.4	0.89 ± 0.01
Second Stage Classification (ccRCC vs. nccRCC) Performance (Mean ± SD ≈)
Method		Acc%	ccRCC/40	nccRCC/30
RC-CAD (Proposed)		89.6 ± 5.0	35 ± 2	28 ± 3
Kunapuli [27]		60.6 ± 2.7	28 ± 1	15 ± 1
Oberai [37]		84.3 ± 3.1	34 ± 1	25 ± 2
Lee [32]	AlexNet	71.7 ± 1.9	31 ± 2	19 ± 2
	GoogleNet	68.0 ± 1.5	32 ± 1	15 ± 1
	ResNet	70.3 ± 2.5	32 ± 0	17 ± 2
	VGGNet	72.6 ± 2.3	33 ± 1	18 ± 1

Hyperparameters: MLP-ANN (optimization function: trainlm, max epochs = 500, hidden layers: hl${}_1$ = 50 nodes, hl${}_2$ = 25 nodes, goal = 0, max validation failure = 6, min gradient = ${10}^{-7}$, training gain ($\mu$): initial $\mu$ = 0.001, $\mu$ decrease factor = 0.1, $\mu$ increase factor = 10, max $\mu$ = 1e${}^{10}$).