Table 2 Comparison of performance among deep learning prediction models integrated with radiomics using concatenation, including ResNet-18, ResNet-34, ResNet-50, DenseNet-121, and Swin Transformer, on the internal validation set, independent test set 1, and independent test set 2

ResNet-18

Internal Validation Set

0.777

0.832

0.723

0.750

0.811

0.837

0.781–0.892

0.746 (0.018)

Independent Test Set 1

0.674

0.767

0.652

0.340

0.923

0.726

0.644–0.807

Independent Test Set 2

0.620

0.798

0.566

0.357

0.902

0.712

0.657–0.768

ResNet-18^a

Internal Validation Set

0.767

0.861

0.673

0.725

0.829

0.813

0.753–0.874

0.754 (0.018)

Independent Test Set 1

0.736

0.698

0.745

0.390

0.913

0.758

0.679–0.838

Independent Test Set 2

0.700

0.677

0.706

0.411

0.878

0.728

0.673–0.782

ResNet-34

Internal Validation Set

0.738

0.713

0.762

0.750

0.726

0.783

0.720–0.847

0.710 (0.019)

Independent Test Set 1

0.740

0.581

0.777

0.379

0.888

0.680

0.586–0.775

Independent Test Set 2

0.636

0.667

0.627

0.351

0.861

0.702

0.647–0.757

ResNet-34^a

Internal Validation Set

0.777

0.871

0.683

0.733

0.841

0.813

0.752–0.874

0.746 (0.018)

Independent Test Set 1

0.753

0.791

0.745

0.420

0.938

0.778

0.699–0.856

Independent Test Set 2

0.638

0.788

0.593

0.370

0.902

0.725

0.672–0.778

ResNet-50

Internal Validation Set

0.757

0.693

0.822

0.795

0.728

0.811

0.752–0.870

0.722 (0.019)

Independent Test Set 1

0.599

0.814

0.549

0.297

0.927

0.698

0.606–0.790

Independent Test Set 2

0.629

0.737

0.596

0.356

0.882

0.677

0.619–0.736

ResNet-50^a

Internal Validation Set

0.733

0.911

0.554

0.672

0.862

0.794

0.732–0.856

0.733 (0.018)

Independent Test Set 1

0.670

0.698

0.663

0.326

0.904

0.714

0.631–0.797

Independent Test Set 2

0.660

0.747

0.633

0.381

0.892

0.738

0.687–0.790

DenseNet-121

Internal Validation Set

0.743

0.911

0.574

0.681

0.866

0.806

0.747–0.866

0.674 (0.020)

Independent Test Set 1

0.656

0.628

0.663

0.303

0.884

0.679

0.593–0.766

Independent Test Set 2

0.540

0.687

0.495

0.292

0.839

0.607

0.543–0.671

Swin Transformer

Internal Validation Set

0.733

0.891

0.574

0.677

0.841

0.775

0.711–0.840

0.742 (0.018)

Independent Test Set 1

0.722

0.628

0.745

0.365

0.895

0.722

0.643–0.800

Independent Test Set 2

0.643

0.828

0.587

0.378

0.919

0.755

0.705–0.805

Swin Transformer^a

Internal Validation Set

0.733

0.832

0.634

0.694

0.790

0.788

0.725–0.850

0.737 (0.018)

Independent Test Set 1

0.727

0.628

0.750

0.370

0.896

0.740

0.661–0.818

Independent Test Set 2

0.646

0.818

0.593

0.379

0.915

0.729

0.677–0.781