You are viewing the site in preview mode

Skip to main content

Table 4 Median and standard deviation of AUROC performance

From: Weakly-supervised deep learning models enable HER2-low prediction from H &E stained slides

 

ADMIL

CLAM

TransMIL

M1

0.87 +0.02

0.90 +0.01

0.88 +0.03

M2

0.85 +0.02

0.85 +0.02

0.85 +0.02

M3

0.84 +0.02

0.87 +0.02

0.88 +0.04

M4

N/A

0.85 +0.01

N/A

M5

0.73 +0.05

0.72 +0.02

0.71 +0.02

M6

N/A

0.78 +0.02

N/A

  1. Median and standard deviation of AUROC performance on the internal test set for ADMIL, CLAM, and TransMIL. ADMIL and TransMIL can only work on binary models. The best performance for each model is highlighted in bold