Development and Validation of Deep Learning Models for the Multiclassification of Reflux Esophagitis Based on the Los Angeles Classification

Table 1

Performance metrics of models and endoscopists.

	Models	Accuracy	Matthew’s correlation coefficient	Cohen’s kappa
Validation dataset
	MobileNet	0.916	0.859	0.820
	ResNet	0.931	0.884	0.850
	Xception	0.938	0.896	0.860
	EfficientNet	0.962	0.936	0.910
	ViT	0.933	0.888	0.850
	ConvMixer	0.950	0.916	0.890
Test dataset
	MobileNet	0.916	0.821	0.780
	ResNet	0.933	0.846	0.810
	Xception	0.936	0.852	0.810
	EfficientNet	0.957	0.884	0.850
	ViT	0.938	0.854	0.820
	ConvMixer	0.943	0.861	0.820
	Junior endoscopist	0.916	0.820	0.780
	Senior endoscopist	0.945	0.864	0.830

The bold figures indicate the highest numeric values.