Table 2 Comparison with baseline models.

From: Combining graph neural network and Mamba to capture local and global tissue spatial relationships in whole slide images

Model

Parameters

TrainTime

InfTime

C-index

Dynamic AUC

MLP

67,905

0.202 ± 0.0369

0.008 ± 0.0006

0.659 ± 0.0189

0.657 ± 0.0348

AttentionMIL

99,202

112.321 ± 11.6783

39.825 ± 2.7891

0.653 ± 0.0211

0.658 ± 0.0539

TransMIL

105,041

135.430 ± 2.1422

54.175 ± 16.4544

0.673 ± 0.0291

0.652 ± 0.0493

MambaMIL

108,738

104.623 ± 1.148

34.758 ± 1.1559

0.675 ± 0.0456

0.669 ± 0.0496

SCMIL

395,139

156.011 ± 15.1807

57.469 ± 13.1659

0.680 ± 0.0287

0.675 ± 0.0382

WiKG

86,658

7.036 ± 0.0490

0.346 ± 0.1055

0.652 ± 0.0271

0.668 ± 0.0127

PatchGCN

82,370

2.613 ± 0.0510

0.447 ± 0.1143

0.616 ± 0.0502

0.634 ± 0.0504

GTP

103,633

0.756 ± 0.0678

0.264 ± 0.1242

0.675 ± 0.0242

0.658 ± 0.0219

Clinical

12

0.132 ± 0.01280

0.003 ± 0.0004

0.608 ± 0.0331

0.622 ± 0.0478

GAT-Mamba

127,425

0.189 ± 0.0736

0.286 ± 0.1551

0.700 ± 0.0228

0.686 ± 0.0281

  1. TrainTime: average training time per epoch, in seconds; InfTime: average inference time across five test sets from cross-validation, in seconds.
Back to article page