Table 1 A variety of metrics computed on the test sets. A threshold of 0.47 was used.

From: A deep learning model to detect pancreatic ductal adenocarcinoma on endoscopic ultrasound-guided fine-needle biopsy

Metrics

Value

Confidence interval

ROC AUC

0.9836

[0.9603–0.9977]

Log loss

0.3419

[0.2949–0.3864]

Accuracy

0.9417

[0.8917–0.975]

MCC

0.8667

[0.7622–0.9473]

f1-score

0.9581

[0.915–0.9827]

Sensitivity (TPR)

0.9302

[0.8602–0.9753]

Specificity (TNR)

0.9706

[0.9091–1]

Precision (PPV)

0.9877

[0.9571–1]

Negative predictive value (NPV)

0.8462

[0.7297–0.9512]

False discovery rate (FDR)

0.0123

[0–0.0429]

Back to article page