Skip to main content
. 2022 Feb 24;17(2):e0264140. doi: 10.1371/journal.pone.0264140

Table 1. Comparison of diagnostic performance for each Convolutional Neural Network (CNN) model.

CNN model Accuracy Sensitivity Specificity Precision F1-Score
ResNet50 0.810 ± 0.046 0.764 ± 0.088 0.889 ± 0.046 0.780 ± 0.059 0.800 ± 0.045
Inception v3 0.821 ± 0.029 0.778 ± 0.087 0.892 ± 0.035 0.786 ± 0.054 0.778 ± 0.054
EfficientNet-b1 0.835 ± 0.054 0.784 ± 0.074 0.896 ± 0.033 0.809 ± 0.082 0.794 ± 0.067
EfficientNet-b2 0.853 ± 0.050 0.822 ± 0.087 0.912 ± 0.034 0.829 ± 0.089 0.822 ± 0.065
EfficientNet-b3 0.842 ± 0.040 0.787 ± 0.090 0.900 ± 0.036 0.818 ± 0.088 0.797 ± 0.068

The values are given as the mean and the standard deviation by 5-fold cross validation.