Table 8 Performance metrics of Resnext50 classification
From: The value of deep neural networks in the pathological classification of thyroid tumors
Classification | Recall | Precision | NPV | Accuracy | Specificity | F1 | AUC | Kappa coefficient |
|---|---|---|---|---|---|---|---|---|
PTC | 94.22% | 90.48% | 94.02% | 92.18% | 90.16% | 92.31% | 0.981 | 0.7508 |
MTC | 47.62% | 90.91% | 95.26% | 95.07% | 99.55% | 62.50% | 0.978 | |
FTC | 80.00% | 44.44% | 98.61% | 92.59% | 93.42% | 57.14% | 0.970 | |
Adenoma | 50.00% | 68.75% | 95.15% | 93.42% | 97.74% | 57.89% | 0.948 | |
Adenomatous goiter | 58.33% | 63.64% | 97.84% | 96.30% | 98.27% | 60.87% | 0.889 | |
Normal thyroid | 92.31% | 92.31% | 97.91% | 96.71% | 97.91% | 92.31% | 0.983 |