Table 2.
Provides an in-depth analysis of performance of the four classes namely glioma, meningioma, pituitary, and no tumor for the three deep learning models for multi-class brain tumor classification task using MRI scan imagery.
| Model | Class | Evaluation metrics | ||||
|---|---|---|---|---|---|---|
| Accuracy (%) |
F1 score (%) |
Precision (%) |
Recall (%) |
Images misclassified | ||
| ResNet-50 | Glioma | 96.64 | 95.05 | 93.50 | 96.64 | 11 |
| Meningioma | 86.67 | 89.98 | 93.56 | 86.67 | 38 | |
| Pituitary | 97.61 | 97.28 | 96.95 | 97.61 | 8 | |
| No tumor | 97.31 | 95.70 | 94.14 | 97.31 | 8 | |
| EfficientNet-B0 | Glioma | 96.64 | 95.93 | 95.22 | 96.64 | 9 |
| Meningioma | 87.72 | 90.74 | 93.98 | 87.72 | 35 | |
| Pituitary | 96.93 | 97.09 | 97.26 | 96.93 | 9 | |
| No tumor | 98.65 | 96.07 | 93.61 | 98.65 | 4 | |
| MobileNet-v2 | Glioma | 95.90 | 94.66 | 93.45 | 95.90 | 11 |
| Meningioma | 86.67 | 89.98 | 93.56 | 86.67 | 38 | |
| Pituitary | 96.93 | 97.09 | 97.26 | 96.93 | 9 | |
| No tumor | 98.99 | 96.24 | 93.63 | 98.99 | 3 | |