Table 5 Performance of radiomic, deep learning, and combined Rad-DL models in quantitative prediction of AMH.

From: Ultrasound radiomics and deep learning for predicting antral follicle count and anti-Müllerian hormone

Model

R2

MAE

MSE

MedAE

EV

Training cohort (n = 210)

 Radiomics

0.520

1.488

4.392

1.126

0.520

 Deep learning

0.950

0.532

0.460

0.431

0.951

 Rad-DL

0.962

0.381

0.344

0.212

0.972

Internal test cohort (n = 91)

 Radiomics

0.323

1.668

4.931

1.315

0.345

 Deep learning

0.591

1.342

2.977

1.026

0.595

 Rad-DL

0.550

1.368

3.279

1.042

0.551

External test cohort (n = 94)

 Radiomics

0.420

1.800

5.729

1.411

0.494

 Deep learning

0.352

1.589

6.405

1.030

0.352

 Rad-DL

0.509

1.553

4.851

1.232

0.523

  1. MAE, mean absolute error; MSE, mean square error; MedAE, median absolute error; EV, explained variance; AMH, anti-Müllerian hormone; Rad, radiomics; DL, deep learning.
Back to article page