Table 1.
Image input model architecture optimization—evaluation metrics with mean and standard deviation with 95% confidence interval.
| w/ | |||||||
|---|---|---|---|---|---|---|---|
| Model | Acc. | AUC | Precision | Recall | |||
| NOR | LAD | LCX | RCA | ||||
| VGG16 | 0.670 ± 0.030 | 1.000 ± 0.000 | 0.873 ± 0.036 | 0.807 ± 0.064 | 0.913 ± 0.024 | 0.720 ± 0.039 | 0.674 ± 0.020 |
| ResNet50V2 | 0.827 ± 0.007 | 1.000 ± 0.000 | 0.927 ± 0.026 | 0.903 ± 0.024 | 0.950 ± 0.000 | 0.836 ± 0.012 | 0.831 ± 0.12 |
| Xception | 0.850 ± 0.023 | 1.000 ± 0.000 | 0.940 ± 0.023 | 0.907 ± 0.033 | 0.963 ± 0.007 | 0.855 ± 0.009 | 0.851 ± 0.011 |
| Inception ResNetV2 |
0.857 ± 0.035 | 1.000 ± 0.000 | 0.957 ± 0.013 | 0.943 ± 0.017 | 0.96 ± 0.011 | 0.847 ± 0.012 | 0.840 ± 0.012 |
| DenseNet121 | 0.843 ± 0.007 | 1.000 ± 0.000 | 0.953 ± 0.007 | 0.920 ± 0.023 | 0.953 ± 0.007 | 0.851 ± 0.014 | 0.831 ± 0.012 |
| InceptionV3 | 0.876 ± 0.025 | 1.000 ± 0.000 | 0.958 ± 0.019 | 0.944 ± 0.024 | 0.970 ± 0.011 | 0879 ± 0.020 | 0.873 ± 0.025 |
| w/o | |||||||
| VGG16 | 0.250 ± 0.000 | 0.500 ± 0.000 | 0.500 ± 0.000 | 0.500 ± 0.000 | 0.500 ± 0.000 | 0.085 ± 0.043 | 0.252 ± 0.004 |
| ResNet50V2 | 0.854 ± 0.013 | 1.000 ± 0.000 | 0.952 ± 0.004 | 0.908 ± 0.010 | 0.966 ± 0.005 | 0.856 ± 0.017 | 0.852 ± 0.015 |
| Xception | 0.856 ± 0.005 | 1.000 ± 0.000 | 0.954 ± 0.021 | 0.928 ± 0.016 | 0.968 ± 0.004 | 0.857 ± 0.006 | 0.855 ± 0.005 |
| Inception ResNetV2 |
0.872 ± 0.010 | 1.000 ± 0.000 | 0.950 ± 0.015 | 0.924 ± 0.017 | 0.976 ± 0.010 | 0.875 ± 0.009 | 0.872 ± 0.010 |
| DenseNet121 | 0.890 ± 0.014 | 1.000 ± 0.000 | 0.978 ± 0.007 | 0.936 ± 0.025 | 0.966 ± 0.012 | 0.893 ± 0.010 | 0.889 ± 0.013 |
| InceptionV3 | 0.900 ± 0.012 | 1.000 ± 0.000 | 0.966 ± 0.010 | 0.948 ± 0.014 | 0.978 ± 0.010 | 0.903 ± 0.011 | 0.899 ± 0.012 |
Acc.: Accuracy, AUC: The area under the ROC curve, NOR: Normal, LAD: Left anterior descending, LCX: Left circumflex artery, RCA: Right coronary artery, w/: with a dense layer, w/o: without a dense layer.