Table 1 Performance comparison between EBVNet and pathologists on MultiCenter-STAD and TCGA-STAD.

From: A deep learning model and human-machine fusion for prediction of EBV-associated gastric cancer from histopathology

Methods

MultiCenter-STAD

TCGA-STAD

Sensitivity

Difference

P

Specificity

Difference

P

Sensitivity

Difference

P

Specificity

Difference

P

EBVNet

0.969

NA

NA

0.759

NA

NA

0.792

NA

NA

0.833

NA

NA

Junior pathologist 1

0.714

−0.255

<0.001

0.850

0.091

<0.001

0.417

−0.375

0.004

0.863

0.030

0.337

Junior pathologist 2

0.704

−0.265

<0.001

0.859

0.100

<0.001

0.417

−0.375

0.004

0.842

0.009

0.871

Senior pathologist 1

0.714

−0.255

<0.001

0.909

0.150

<0.001

0.417

−0.375

0.012

0.932

0.099

<0.001

Senior pathologist 2

0.633

−0.336

<0.001

0.934

0.175

<0.001

0.458

−0.334

0.008

0.893

0.060

0.016

Expert pathologist 1

0.684

−0.285

<0.001

0.959

0.200

<0.001

0.542

−0.250

0.070

0.923

0.090

<0.001

Expert pathologist 2

0.714

−0.255

<0.001

0.918

0.159

<0.001

0.542

−0.250

0.070

0.932

0.099

<0.001

  1. Difference indicated the difference between each pathologist and EBVNet. The data have been provided in the Source Data file.
  2. MultiCenter-STAD external dataset from multiple medical center, TCGA-STAD external dataset from The Cancer Genome Atlas, NA not applicable.
Back to article page