Table 1 ResNet-50, DenseNet-121, EfficientNet-B6, and ViT 5-fold cross validation results for classification.

From: Multi-institutional validation of AI models for classifying urothelial neoplasms in digital pathology

Model

Accuracy

(95% CI)

Sensitivity

(95% CI)

Specificity

(95% CI)

F1-score

(95% CI)

ResNet-50

0.887

(0.884–0.889)

0.885

(0.882–0.888)

0.943

(0.942–0.944)

0.886

(0.884–0.888)

DenseNet-121

0.905

(0.903–0.907)

0.902

(0.898–0.906)

0.952

(0.951–0.953)

0.901

(0.896–0.906)

EfficientNet-B6

0.913

(0.907-920)

0.909

(0.904–0.914)

0.956

(0.953–0.960)

0.906

(0.901–0.911)

ViT

0.722

(0.696–0.749)

0.718

(0.706–0.730)

0.861

(0.848–0.874)

0.720

(0.695–0.745)

Back to article page