Table 1 Classification accuracy and F1 scores of Inception v3, ResNet101 and DenseNet201 convolutional neural networks on calcified plaques with motion artifacts of four densities
Plaque density | Inception v3 | ResNet101 | DenseNet201 | |||
|---|---|---|---|---|---|---|
Accuracy | F1 score | Accuracy | F1 score | Accuracy | F1 score | |
High | 88.8 ± 2.3% | 0.917 ± 0.024 | 90.2 ± 2.8% | 0.922 ± 0.027 | 89.3 ± 2.9% | 0.919 ± 0.023 |
Medium-1 | 88.0 ± 3.0% | 0.901 ± 0.022 | 87.1 ± 2.3% | 0.896 ± 0.020 | 87.7 ± 2.9% | 0.897 ± 0.021 |
Medium-2 | 90.7 ± 2.5% | 0.939 ± 0.024 | 92.9 ± 2.9% | 0.942 ± 0.028 | 90.7 ± 2.0% | 0.937 ± 0.018 |
Low | 93.3 ± 1.6% | 0.950 ± 0.012 | 92.4 ± 2.8% | 0.947 ± 0.020 | 92.7 ± 2.8% | 0.945 ± 0.019 |