Table 3 Classification accuracy and 95% confidence intervals for machine learning models on the LR-M test set with average radiologist performance comparison on the 1st test-set segmentation.

From: Automated machine learning for differentiation of hepatocellular carcinoma from intrahepatic cholangiocarcinoma on multiphasic MRI

 

ROC AUC

PR AUC

Accuracy

P value

Sensitivity

P value

Specificity

P value

Kappa

Manual

0.68

0.91

0.71 (0.51–0.85)

0.20

0.79 (0.56–0.92)

1.00

0.40 (0.12–0.77)

0.03

0.18

TPOT

0.60

0.89

0.63 (0.43–0.79)

0.03

0.68 (0.45–0.85)

0.08

0.40 (0.12–0.77)

0.03

0.07

Radiologist 1

NA

NA

0.79 (0.59–0.91)

0.44

0.79 (0.56–0.92)

1.00

0.80 (0.36–0.98)

0.56

0.48

Radiologist 2

NA

NA

0.79 (0.59–0.91)

0.44

0.79 (0.56–0.92)

1.00

0.80 (0.36–0.98)

0.56

0.48

Radiologist 3

NA

NA

0.83 (0.63–0.94)

0.80

0.79 (0.56–0.92)

1.00

1.00 (0.51–1.00)

1.00

0.61

Radiologist 4

NA

NA

0.83 (0.63–0.94)

0.80

0.84 (0.61–0.95)

1.00

0.80 (0.36–0.98)

0.56

0.56

Mean radiologist

NA

NA

0.81 (0.61–0.92)

1.00

0.80 (0.57–0.93)

1.00

0.85 (0.40–1.00)

1.00

NA