Table 8.

Performance results of CNN-RF hybrid models based on the fused CNN models for early diagnosis and differentiation of skin cancer and other skin lesions.

Fusion Features	Classifier	Type of Lesion	AUC %	Sensitivity %	Accuracy %	Precision %	Specificity %
AlexNet-GoogLeNet	RF	Scc	88.5	77.4	77	80.2	99.5
		Akiec	90.5	85.2	85.1	82.2	98.8
		Bcc	97.5	96.3	95.8	95.6	99.2
		Bkl	96.2	95.7	95.8	93.8	99.6
		Df	78.5	60.4	60.4	69	100
		Mel	97.5	96.1	96.1	93.1	97.9
		Nv	98.3	97.8	97.6	98.8	98.2
		Vasc	86.5	71.3	70.6	76.6	100
GoogLeNet-VGG16	RF	Scc	92.1	87	86.5	84.5	99.6
		Akiec	91.4	83.1	82.8	86.2	99.5
		Bcc	97.2	96.9	96.7	93.9	98.7
		Bkl	96.4	94.2	94.1	94.5	99.1
		Df	79.6	69.3	68.8	75	100
		Mel	97.5	96.2	95.6	93.4	99.3
		Nv	98.5	97.3	97.2	98.4	98.2
		Vasc	86.5	76.1	76.5	76.5	100
AlexNet-VGG16	RF	Scc	85.2	82.4	81.7	79.8	99.1
		Akiec	82.5	80.1	79.9	83.7	99.4
		Bcc	98.1	96.3	96.1	94.4	98.8
		Bkl	96.4	93.8	93.5	95.5	99.1
		Df	81.6	73.2	72.9	64.8	100
		Mel	97.1	93.4	93.4	90.9	97.8
		Nv	98.1	97.1	96.5	97.6	98.4
		Vasc	88.1	78.2	78.4	76.9	100
AlexNet-GoogLeNet-VGG16	RF	Scc	86.2	81.4	81	86.4	99.5
		Akiec	88.3	87.8	87.9	81.8	99.1
		Bcc	98.2	96.2	95.8	95.8	98.8
		Bkl	97.5	95.4	96	94.7	99
		Df	75.1	65.3	64.6	66	100
		Mel	98.2	97.2	96.6	93.7	99.4
		Nv	98.9	98.1	97.7	99.1	98.6
		Vasc	80.2	70.9	70.6	76.6	100