Skip to main content

Table 2 Diagnostic performance of the VGGNet-16 model and gynecologists in the five-category classification task

From: Deep learning model for classifying endometrial lesions

Category Sensitivity (%) Specificity (%) Precision (%) F1-score (%) AUC Accuracy (%)
VGGNet-16
 EH 84.0 92.5 73.7 78.5 0.926 80.8
 AH 68.0 95.5 79.1 73.1 0.916
 EC 78.0 96.5 84.8 81.3 0.952
 EP 94.0 95.0 82.5 87.9 0.981
 SM 80.0 96.5 85.1 82.5 0.959
Gynecologist 1
 EH 70.0 90.0 63.6 66.7 0.800 72.8
 AH 58.0 92.5 65.9 61.7 0.753
 EC 74.0 90.0 64.9 69.2 0.820
 EP 86.0 95.0 81.1 83.5 0.905
 SM 76.0 98.5 92.7 83.5 0.873
Gynecologist 2
 EH 64.0 94.5 74.4 68.8 0.792 69.2
 AH 54.0 90.0 57.4 55.7 0.720
 EC 68.0 92.5 69.4 68.7 0.803
 EP 90.0 87.0 63.4 74.4 0.885
 SM 70.0 97.5 87.5 77.8 0.838
Gynecologist 3
 EH 52.0 95.0 72.2 60.5 0.735 64.4
 AH 54.0 87.0 50.9 52.4 0.705
 EC 66.0 93.0 70.2 68.0 0.795
 EP 80.0 87.0 60.6 69.0 0.835
 SM 70.0 93.5 72.9 71.4 0.818
  1. AH atypical hyperplasia, AUC area under the receiver operating characteristic (ROC) curve, EC endometrial cancer, EH endometrial hyperplasia without atypia, EP endometrial polyp, SM submucous myoma