Table 7.

A comparison between CoMB-Deep and related studies using the same dataset.

Classification category
	Method	Sensitivity (%)	Precision (%)	Specificity (%)	Accuracy (%)
Das et al. (2019)	HOG, GLCM, Tamura, and LBP features + GRLN + SVM	100	100	100	100
Das et al. (2020c)	AlexNet + SVM VGG-16 + SVM	–	–	–	99.44 99.62
Das et al. (2018b)	(Shape + Color) features + PCA + SVM	100	100	100	100
Das et al. (2020c)	AlexNet sVGG-16	–	–	–	98.5 98.12
Das et al. (2020a)	HOG, GLCM, Tamura, and LBP features + GRLN + MANOVA + SVM	100	100	100	100
Proposed CoMB-Deep	Inception-ResNet + DWT + Information gain + LSTM	100	100	100	100
Classification (multi-class category)
		Sensitivity (%)	Precision (%)	Specificity (%)	Accuracy (%)
Das et al. (2018b)	(Shape + Color) features + PCA + SVM	–	–	–	84.9
Das et al. (2020a)	HOG, GLCM, Tamura, and LBP features + GRLN + MANOVA + SVM	72	66.6	–	65.21
Das et al. (2020c)	AlexNet VGG-16 s	–	–	–	79.33 65.4
Das et al. (2020b)	GLCM + Tamura + LBP + GRLN + SVM	91.3	91.3	97	91.3
Das et al. (2020b)	GLCM + Tamura + LBP + GRLN + PCA + SVM	–	–	–	96.7
Das et al. (2020c)	AlexNet + SVM VGG-16 + SVM	–	–	–	93.21 93.38
Proposed CoMB-Deep	Deep features of (DenseNet-201 + ShuffleNet) + Relief-F + Bi-LSTM	98.1	98.1	99.3	98.05
	Deep features of (DenseNet-201 + Inception + Resnet-50 + Darknet-53 + MobileNet + ShuffleNet + SqueezeNet + NasNetMobile) + Relief-F + Bi-LSTM	99.8	99.4	99.4	99.35