Table 2.
Methodological characteristics of the included studies
| Citation | Modality | Annotation | Number of readers per case | Reader agreement after annotation | Feature extraction | Number of extracted features | Number of selected features | Final classifier | AUC |
|---|---|---|---|---|---|---|---|---|---|
| Xu et al [12] | MRI | Manual; 3D | 2 | Yes | Radiomics | 63 | 30 | Support vector machine | / |
| Garapati et al [13] | CT | Automatic; 3D | 1 | No | Radiomics | 91 | Lack information | Random forest | / |
| Xu and Zhang et al [14] | MRI | Manual; 2D | 1 | No | Radiomics | 1104 | 19 | Support vector machine | 0.9857 |
| Zheng et al [15] | MRI | Semi-automatic; 3D | 1 | Yes | Radiomics | 2062 | 23 | LASSO | / |
| Xu and Yao et al [16] | MRI | Manual; 3D | 1 | No | Radiomics | 156 | 21 | Random forest | 0.907 |
| Wang et al [17] | MRI | Manual; 2D | 2 | Yes | Radiomics | 1404 | 36 | Logistic regression | 0.672 |
| Zhou et al [18] | CT | Semi-automatic; 3D | 2 | No | Radiomics | 1223 | 6 | Support vector machine; logistic regression | 0.782 |
| Zhang et al [19] | CT | Semi-automatic; 3D | 2 | No | Deep learning | Lack information | Lack information | FGP-Net | 0.791 |
| Zheng and Xu et al [20] | MRI | Manual; 3D | 1 | No | Radiomics | 2436 | 21 | LASSO | 0.906 |
| Yang et al [21] | CT | Manual; 2D | 2 | Yes | Deep learning | 49 | 49 | Neural network | 0.998 |
| Gao et al [22] | US | Manual; 2D | 2 | Yes | Radiomics | 5936 | 30 | Naïve Bayes | 0.84 |
| Chen et al [23] | CT | Manual; 3D | Lack information | No | Both | 1735; 2048 | 30 | Logistic regression | 0.884 |
| Zou et al [24] | MRI | Automatic; 2D | 2 | Yes | Deep learning | 2048 | 2048 | Inception V3 | 0.856 |
| Zhang and Li et al [25] | MRI | Manual; 3D | 2 | Yes | Radiomics | 23,688 | 9 | Logistic regression | 0.931 |
| Cui et al [26] | CT | Manual; 3D | 1 | No | Radiomics | 102 | 8 | AdaBoost | 0.894 |
| Zhang and Wu et al [27] | CT | Semi-automatic; 3D | 2 | No | Radiomics | 1218 | 8 | Logistic regression | 0.784 |
| Liu et al [28] | MRI | Manual; 2D | 1 | No | Radiomics | 4128 | 28 | LASSO | / |
| Sarkar et al [29] | CT | Manual; 3D | 1 | Lack information | Deep learning | 2048–9216 | Lack information | Linear discriminant analysis | / |
| Li et al [30] | MRI | Manual; 3D | 1 | No | Both | 851; Lack information | Lack information | Neural network | 0.932 |
| Wang and Li et al [31] | MRI | Manual; 3D | 2 | No | Radiomics | 1070 | 6 | Logistic regression | 0.711 |
| Li and Cao et al [32] | MRI | Manual; 3D | 1 | No | Deep learning | 2048 | 2048 | ResNet | 0.861 |
US ultrasound, LASSO least absolute shrinkage and selection operator