Table 3.

Performance of the CNN1-HA model in polyp subgroup.

Datasets	Support images	Precision	Recall	F1-score	Accuracy	AUC
Adenoma	307	0.883 (0.847, 0.919)	0.863 (0.825, 0.901)	0.873 (0.836, 0.910)	0.925 (0.896, 0.954)	0.975 (0.963, 0.987)
Hyperplastic	716	0.942 (0.925, 0.959)	0.950 (0.934, 0.966)	0.946 (0.929, 0.963)	0.924 (0.905, 0.943)
All	1023	0.925 (0.909, 0.941)	0.924 (0.908, 0.940)	0.924 (0.908, 0.940)	0.924 (0.908, 0.940)

95% confidence intervals are included in brackets.

CNN1-HA model represents the ResNet-50 classifying the hyperplastic and adenomatous polyp.

Abbreviations: AUC, area under the curve.