Hybrid Techniques for Diagnosing Endoscopy Images for Early Detection of Gastrointestinal Disease Based on Fusion Features

Table 2

Results of CNN + SVM systems for diagnosing endoscopy images of the gastrointestinal disease dataset.

System	Disease types	Accuracy (%)	Sensitivity (%)	Precision (%)	Specificity (%)	AUC (%)
VGG-16 + SVM	Dyed_lifted_polyps	92.50	93.26	96.90	100	98.52
	Dyed_resection_margins	96.50	96.32	95.50	99.35	99.12
	Esophagitis	95.50	95.34	93.60	99.42	99.50
	Normal_cecum	95.00	94.84	95.00	98.78	98.80
	Normal_pylorus	97.00	97.28	96.50	99.10	99.15
	Normal_z_line	98.50	97.92	97.50	100	99.66
	Polyps	95.50	95.10	96.00	98.95	98.95
	Ulcerative_colitis	94.00	93.85	93.50	99.22	98.45
	Average ratio	95.60	95.49	95.56	99.35	99.02
DenseNet-121 + SVM	Dyed_lifted_polyps	97	97.12	97.50	99.59	98.17
	Dyed_resection_margins	99	98.84	97.10	99.87	99.81
	Esophagitis	96.5	96.35	96.50	99.11	98.25
	Normal_cecum	95.5	94.91	97.40	100	97.99
	Normal_pylorus	96.5	95.68	97.00	99.69	98.69
	Normal_z_line	97.5	97.26	97.00	100	99.10
	Polyps	96.00	96.09	96.50	98.78	97.84
	Ulcerative_colitis	93.50	94.41	92.60	98.87	98.74
	Average ratio	96.40	96.33	96.45	99.49	98.57