Research Article
Handcrafted and Deep Learning-Based Radiomic Models Can Distinguish GBM from Brain Metastasis
Table 2
The performance comparison of multimodality and single-modality models.
| Models | Training AUC | Cross-validation mean AUC | Test AUC | Accuracy | Sensitivity | Specificity |
| T1CE-HCR | 0.99 | 0.94 | 0.93 | 0.82 | 0.70 | 0.93 | T1CE-HCR+DLR | 0.99 | 0.95 | 0.97 | 0.85 | 0.84 | 0.93 | T1WI-HCR | 1.00 | 0.90 | 0.86 | 0.71 | 0.75 | 0.78 | T1WI-HCR+DLR | 0.99 | 0.91 | 0.87 | 0.76 | 0.75 | 0.83 | T2WI-HCR | 1.00 | 0.95 | 0.76 | 0.73 | 0.85 | 0.66 | T2WI-HCR+DLR | 1.00 | 0.96 | 0.80 | 0.78 | 0.80 | 0.83 | Multimodality-HCR | 1.00 | 0.92 | 0.81 | 0.71 | 0.70 | 0.82 | Multimodality-HCR + DLR | 1.00 | 0.96 | 0.84 | 0.75 | 0.71 | 0.91 |
|
|
The bold values represent the highest accuracy and specificity.
|