Table 2 The performance of ten ML models for recognizing CAS in the training set, internal validation set, and external validation set

Training set

auROC (95% CI)

0.855 (0.845–0.865)

0.878 (0.868–0.888)

0.856 (0.845–0.866)

0.847 (0.836–0.858)

0.837 (0.825–0.848)

0.921 (0.914–0.928)

0.868 (0.858–0.878)

0.822 (0.811–0.834)

0.875 (0.865–0.884)

0.908 (0.900–0.916)

auPR (95% CI)

0.849 (0.835–0.862)

0.852 (0.838–0.865)

0.851 (0.837–0.864)

0.833 (0.818–0.849)

0.812 (0.794–0.829)

0.925 (0.917–0.933)

0.865 (0.852–0.877)

0.777 (0.757–0.798)

0.871(0.859–0.883)

0.903 (0.892–0.913)

Threshold^a

0.547

0.594

0.517

0.539

0.523

0.662

0.57

0.51

0.702

0.652

Sensitivity

0.77

0.814

0.792

0.82

0.814

0.831

0.806

0.801

0.858

0.841

Specificity

0.776

0.779

0.725

0.719

0.708

0.831

0.764

0.709

0.844

0.811

PPV

0.766

0.787

0.744

0.745

0.762

0.813

0.782

0.766

0.833

0.817

NPV

0.776

0.808

0.773

0.798

0.757

0.848

0.782

0.718

0.865

0.836

PLR

3.444

3.692

2.878

2.921

2.792

4.931

3.416

2.754

5.49

4.455

NLR

0.296

0.238

0.287

0.25

0.262

0.203

0.254

0.28

0.168

0.196

Internal Validation set

auROC (95% CI)

0.861(0.841–0.881)

0.800(0.777–0.824)

0.852(0.832–0.872)

0.846(0.824–0.867)

0.835(0.812–0.857)

0.849(0.828–0.870)

0.860(0.839–0.880)

0.829(0.805–0.852)

0.855(0.835–0.876)

0.817(0.794–0.840)

auPR (95% CI)

0.864(0.842–0.885)

0.757(0.725–0.791)

0.857(0.834–0.880)

0.842(0.816–0.867)

0.816(0.785–0.847)

0.826(0.795–0.861)

0.860(0.836–0.883)

0.799(0.766–0.834)

0.857(0.834–0.880)

0.790(0.759–0.823)

Sensitivity

0.85

0.779

0.855

0.867

0.855

0.838

0.84

0.838

0.811

0.789

Specificity

0.722

0.7

0.646

0.693

0.7

0.727

0.729

0.685

0.746

0.749

PPV

0.757

0.72

0.724

0.74

0.759

0.751

0.762

0.769

0.753

0.738

NPV

0.804

0.761

0.759

0.83

0.782

0.812

0.797

0.73

0.797

0.78

PLR

3.057

2.596

2.418

2.822

2.85

3.067

3.104

2.663

3.196

3.139

NLR

0.208

0.316

0.224

0.192

0.207

0.223

0.219

0.237

0.253

0.282

External Validation set

auROC (95% CI)

0.853 (0.839–0.866)

0.800 (0.783–0.815)

0.841 (0.827–0.854)

0.847 (0.833–0.859)

0.833 (0.818–0.846)

0.847 (0.833–0.860)

0.851 (0.837–0.863)

0.812 (0.796–0.826)

0.844 (0.829–0.856)

0.799 (0.783–0.814)

auPR (95% CI)

0.844 (0.824–0.860)

0.750 (0.726–0.772)

0.828 (0.808–0.846)

0.836 (0.816–0.854)

0.816 (0.794–0.836)

0.829 (0.809–0.848)

0.835 (0.813–0.853)

0.767 (0.742–0.793)

0.824 (0.801–0.843)

0.761 (0.738–0.785)

Sensitivity

0.782

0.82

0.811

0.81

0.811

0.799

0.794

0.838

0.81

0.781

Specificity

0.781

0.651

0.695

0.735

0.719

0.756

0.765

0.652

0.728

0.713

PPV

0.748

0.7

0.71

0.722

0.741

0.731

0.75

0.736

0.728

0.718

NPV

0.813

0.784

0.798

0.827

0.787

0.805

0.803

0.725

0.808

0.775

PLR

3.574

2.348

2.658

3.053

2.882

3.278

3.383

2.405

2.973

2.716

NLR

0.279

0.277

0.271

0.259

0.264

0.266

0.269

0.249

0.262

0.307