Research Article
Thyroid Diagnosis from SPECT Images Using Convolutional Neural Network with Optimization
Table 3
Subacute disease class performance comparison of different methods (percent).
| Network | DenseNet121 | ResNet101 | InceptionV3 | VGG19 | MVGG | GoogleNet | SDAE | Proposed |
| Recall | 98.17 | 97.33 | 89.17 | 94.00 | 92.50 | 92.83 | 95.33 | 100.00 | Precision | 96.88 | 96.21 | 93.04 | 94.31 | 90.54 | 96.20 | 93.77 | 98.68 | Accuracy | 98.75 | 98.38 | 95.63 | 97.08 | 95.71 | 97.29 | 97.25 | 99.67 | Specificity | 98.94 | 98.72 | 97.78 | 98.11 | 96.78 | 98.78 | 97.89 | 99.56 | F1 score | 97.52 | 96.77 | 91.06 | 94.16 | 91.51 | 94.49 | 94.55 | 99.34 |
|
|