ColoRectalCADx: Expeditious Recognition of Colorectal Cancer with Integrated Convolutional Neural Networks and Visual Explanations Using Mixed Dataset Evidence

Table 10

Comparison of precision and support of mixed dataset classes.

Classes		High-performance CNN models
		DenseNet-201	ADaDR-22	Support
		Precision	Precision
0	bbps-0-1	0.97	0.92	198
1	bbps-2-3	0.99	0.96	345
2	Cecum	0.9	0.72	603
3	Dyed-lifted-polyps	0.55	0.58	601
4	Dyed-resection-margins	0.77	0.79	597
5	Esophagitis-a	0.43	0	421
6	Non_polyps	0.97	0.84	257
7	Polyps	0.96	0.68	368
8	Pylorus	0.89	0.9	600
9	Retroflex-rectum	0.94	1	117
10	Retroflex-stomach	0.98	0.97	230
11	Ulcerative-colitis-grade-0-1	0.34	0	311
12	Ulcerative-colitis-grade-2	0.37	0.3	133
13	-line	0.63	0.58	580