Table 1 The c-index of machine learning models in papillary thyroid carcinoma (PTC) prediction applied to different subgroups and different variables used for model construction

LNM

Model

 LR(Logistic regression)

23,903

57,386

0.787(0.772–0.802)

5614

14,556

0.784(0.765–0.802)

 LASSO

357

1113

0.739(0.691–0.787)

177

588

0.695(0.549–0.841)

RF(Random Forest)

2536

5082

0.745(0.671–0.818)

215

414

0.743(0.700-0.787)

 GBM(Gradient Boosting Machine)

2133

4601

0.703(0.629–0.777)

32

118

0.909(0.849–0.969)

 SVM(Support Vector Machine)

1659

2765

0.790(0.767–0.814)

215

414

0.723(0.678–0.767)

 ANN(Artificial Neural Network)

2311

4036

0.721(0.685–0.757)

215

414

0.722(0.677–0.766)

 KNN(K-Nearest Neighbor)

1360

2861

0.696(0.567–0.824)

DT(Decision Tree)

2311

4036

0.643(0.585-0.700)

215

414

0.593(0.544–0.642)

 BN(Batch Normalization)

524

950

0.748(0.720–0.776)

215

414

0.709(0.664–0.791)

 DL(Deep Learning)

225

1046

0.670(0.631–0.709)

Overall

37,319

83,876

0.762(0.747–0.777)

6898

17,332

0.773(0.754–0.791)

LR Variable

Clinical characteristics

19,131

51,893

0.784(0.769-0.800)

4711

11,746

0.778(0.758–0.797)

Radiomics

308

591

0.857(0.800-0.914)

145

359

0.870(0.809–0.931)

Clinical characteristics + Genomics

1137

3768

0.792(0.692–0.892)

379

1070

0.821(0.637-1.000)

Rdiomics + Ultrasonics

55

600

0.946(0.910–0.981)

31

286

0.914(0.841–0.987)

Ultrasonics + Clinical characteristics

131

300

0.905(0.843–0.966)

255

420

0.881(0.853–0.910)

Radiomics + Clinical characteristics

318

637

0.742(0.686–0.797)

140

255

0.672(0.639–0.706)

Central LNM

Model

LR(Logistic regression)

17,021

44,245

0.774(0.759–0.789)

5292

12,511

0.771(0.750–0.791)

LASSO

160

439

0.780(0.735–0.825)

77

220

0.770(0.703–0.837)

RF(Random Forest)

1401

3267

0.725(0.616–0.835)

215

414

0.743(0.700-0.787)

GBM(Gradient Boosting Machine)

998

2786

0.700(0.604–0.795)

32

118

0.909(0.849–0.969

SVM(Support Vector Machine)

524

950

0.776(0.749–0.803)

215

414

0.723(0.678–0.767)

ANN(Artificial Neural Network)

1176

2221

0.739(0.721–0.757)

215

414

0.722(0.677–0.766)

KNN(K-Nearest Neighbor)

225

1046

0.629(0.589–0.669)

DT(Decision Tree)

1176

2221

0.648(0.545–0.752)

215

414

0.593(0.544–0.642)

BN(Batch Normalization)

524

950

0.748(0.720–0.776)

215

414

0.709(0.664–0.754)

DL(Deep Learning)

225

1046

0.670(0.631–0.709)

Overall

23,430

59,171

0.762(0.747–0.777)

6476

14,919

0.762(0.747–0.777)

LR Variable

Clinical characteristics

15,421

40,850

0.773(0.756–0.789)

4533

10,534

0.775(0.754–0.795)

Radiomics

218

311

0.830(0.785–0.875)

81

132

0.792(0.715–0.869)

Clinical characteristics + Genomics

960

2570

0.818(0.700-0.937)

379

1070

0.821(0.637-1.000)

Rdiomics + Clinical characteristics

318

637

0.742(0.686–0.797)

140

255

0.672(0.639–0.706)

Ultrasonics + Clinical characteristics

104

300

0.774(0.759–0.789)

159

420

0.871(0.836–0.907)

Lateral Cervical LNM

Model

LR(Logistic regression)

6882

13,141

0.826(0.792–0.861)

322

2045

0.848(0.813–0.883)

LASSO

197

674

0.712(0.670–0.755)

100

368

0.621(0.560–0.682)

RF(Random Forest)

1135

1815

0.800(0.784–0.816)

GBM(Gradient Boosting Machine)

1135

1815

0.720(0.701–0.739)

SVM(Support Vector Machine)

1135

1815

0.800(0.784–0.816)

ANN(Artificial Neural Network)

1135

1815

0.690(0.671–0.709)

KNN(K-Nearest Neighbor)

1135

1815

0.760(0.742–0.778)

DT(Decision Tree)

1135

1815

0.630(0.609–0.651)

Overall

13,889

24,705

0.803(0.773–0.834)

422

2413

0.829(0.779–0.879)

LR Variable

Clinical characteristics

3710

12,071

0.818(0.781–0.855)

178

1112

0.800(0.765–0.835)

Radiomics

90

280

0.899(0.862–0.936)

64

227

0.908(0.867–0.949)

Clinical characteristics + Genomics

177

1198

0.714(0.674–0.754)

Rdiomics + Ultrasonics

55

600

0.946(0.910–0.981)

31

286

0.914(0.841–0.987)

Ultrasonics + Clinical characteristics

27

300

0.938(0.887–0.989)

49

420

0.901(0.852–0.949)