Table 2 The performance comparison of different models in the internal and external testing set.

From: Multiparameter MRI-based model integrating radiomics and deep learning for preoperative staging of laryngeal squamous cell carcinoma

Model

AUC (95% CI)

F1-Score

Accuracy

Specificity

Sensitivity

PPV

NPV

Internal testing set

 Radiomics-CE-T1WI

0.813 (95% CI 0.726–0.901)

0.763

0.769

0.720

0.829

0.708

0.837

 Radiomics-T1WI

0.807 (95% CI 0.748–0.850)

0.725

0.791

0.887

0.658

0.806

0.783

 Radiomics-T2WI

0.815 (95% CI 0.724–0.905)

0.752

0.791

0.878

0.690

0.828

0.768

 Radiomics-CE-T1WI + T1WI

0.821 (95% CI 0.730–0.913)

0.761

0.813

0.887

0.711

0.818

0.810

 Radiomics-CE-T1WI + T2WI

0.850 (95% CI 0.773–0.928)

0.843

0.835

0.947

0.755

0.952

0.735

 Radiomics-T1WI + T2WI

0.849 (95% CI 0.763–0.934)

0.762

0.780

0.736

0.842

0.696

0.867

 Radiomics-ALL

0.857 (95% CI 0.778–0.936)

0.762

0.780

0.830

0.727

0.800

0.765

 DL model

0.821 (95% CI 0.736–0.906)

0.705

0.714

0.829

0.620

0.816

0.642

 Combined model

0.913 (95% CI 0.852–0.974)

0.876

0.879

0.911

0.848

0.907

0.854

External testing set

 Radiomics-CE-T1WI

0.797 (95% CI 0.710–0.885)

0.795

0.742

0.511

0.942

0.690

0.885

 Radiomics-T1WI

0.776 (95% CI 0.684–0.868)

0.764

0.722

0.578

0.846

0.698

0.765

 Radiomics-T2WI

0.793 (95% CI 0.704–0.881)

0.762

0.742

0.711

0.769

0.755

0.727

 Radiomics-CE-T1WI + T1WI

0.802 (95% CI 0.715–0.889)

0.701

0.711

0.533

0.865

0.592

0.774

 Radiomics-CE-T1WI + T2WI

0.812 (95% CI 0.728–0.897)

0.807

0.775

0.644

0.885

0.741

0.829

 Radiomics-T1WI + T2WI

0.821 (95% CI 0.738–0.904)

0.789

0.784

0.689

0.865

0.763

0.816

 Radiomics-ALL

0.835 (95% CI 0.756–0.915)

0.842

0.814

0.689

0.923

0.774

0.886

 DL model

0.811 (95% CI 0.726–0.896)

0.745

0.732

0.733

0.730

0.760

0.702

 Combined model

0.877 (95% CI 0.807–0.946)

0.839

0.814

0.711

0.904

0.783

0.865

 Doctor 1

0.716 (95% CI 0.614–0.816)

0.708

0.714

0.778

0.654

0.773

0.660

 Doctor 2

0.717 (95% CI 0.616–0.818)

0.703

0.714

0.800

0.635

0.786

0.655

  1. DL, deep learning; AUC, area under the curve; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging; CE-T1WI, contrast-enhanced T1-weighted imaging.
Back to article page