Table 1.
Summary of relevant literature on glioma grading and the corresponding methodologies used. For each study, we report details about the database used, sample size, MRI modality, data spitting strategy, dimensionality, number of slices used, categories considered, and the reported performance.
| Reference | Database | Sample Size | MRI Modality | Split | Dimensionality | Slices | Classes | Performance |
|---|---|---|---|---|---|---|---|---|
| Yang et al. [9] | Private | 113 | T1ce | Patient | 2D | ≥80% tumor | LGG/HGG | OA = 0.945 |
| AUC = 0.968 | ||||||||
| Pereira et al. [10] | BraTS2017 | 285 | T1, T1ce, T2, FLAIR | Patient-level | 3D | All | LGG/HGG | OA = 0.9298 |
| AUC = 0.9841 | ||||||||
| Recall LGG = 0.867 | ||||||||
| Recall HGG = 0.952 | ||||||||
| Banerjee et al. [11] | BraTS2017 | 746 | T1, T1ce, T2, FLAIR | Patient | 2D | largest slice ± 10 | LGG/HGG | OA = 0.970 |
| TCGA | Recall LGG = 0.96 | |||||||
| Recall HGG = 0.943 | ||||||||
| Anaraki et al. [12] | IXI | 1288 | T1ce | Slice | 2D | tumor visible | Normal/G.2/G.3/G.4 | OA = 0.909 |
| REMBRANDT | Recall Normal = 0.998 | |||||||
| TCGA | Recall G.2 = 0.884 | |||||||
| private | Recall G.3 = 0.864 | |||||||
| Recall G.4 = 0.974 | ||||||||
| Tandel et al. [13] | REMBRANDT | 130 | T2 | Slice | 2D | tumor visible | Normal/LGG/HGG | OA = 0.960 |
| AUC = 0.990 | ||||||||
| Zhuge et al. [14] | Brats2018 | 315 | T1ce, T2, FLAIR | Patient | 2D | largest slice | LGG/HGG | OA = 0.963 |
| TCGA | Recall HGG = 0.935 | |||||||
| Recall LGG = 0.972 | ||||||||
| Patient | 3D | All | LGG/HGG | OA = 0.971 | ||||
| Recall HGG = 0.947 | ||||||||
| Recall LGG = 0.968 | ||||||||
| Ayadi et al. [16] | REMBRANDT | 130 | Slice | 2D | tumor visible | Normal/LGG/HGG | OA = 0.972 | |
| Recall Normal = 1 | ||||||||
| Recall LGG = 0.984 | ||||||||
| Recall HGG = 0.860 | ||||||||
| Radiopaedia | 121 | Slice | 2D | tumor visible | G.1/G.2/G.3/G.4 | OA = 0.937 | ||
| Recall G.1 = 0.901 | ||||||||
| Recall G.2 = 0.957 | ||||||||
| Recall G.3 = 0.908 | ||||||||
| Recall G.4 = 0.982 | ||||||||
| Ding et al. [17] | Private | 151 | T1ce | Patient | 2D | largest + 2 adjacent | LGG/HGG | OA= 0.800 |
| TCGA | AUC = 0.898 | |||||||
| Recall LGG = 0.760 | ||||||||
| Recall HGG = 0.840 | ||||||||
| van der Voort et al. [18] | EGD | 1748 | T1, T1ce, T2, FLAIR | Patient | 3D | All | LGG/HGG | OA = 0.84 |
| REMBRANDT | AUC = 0.91 | |||||||
| CPTAC-GBM | Recall LGG = 0.72 | |||||||
| TCGA | Recall HGG = 0.93 | |||||||
| BraTS2019 | G.2/G.3/G.4 | OA = 0.71 | ||||||
| Ivy GAP | AUC = 0.81 | |||||||
| Private | Recall G.2 = 0.75 | |||||||
| Recall G.3 = 0.17 | ||||||||
| Recall G.4 = 0.95 |
OA = Overall Accuracy, AUC = Area Under the (ROC) Curve.