Table 4.
Evaluation metric and outcome performance of the method with either the best or featured model.
| Articles | Model and Approach | Evaluation Metrics and Outcomes | |||||
|---|---|---|---|---|---|---|---|
| Acc (%) | Sn/Rc (%) | Sp (%) | PPV/Pc (%) | NPV (%) | AUC (%) | ||
| Hu et al. [44] | B-mode + SWE (1.0 mm offset) ResNet18 | 86.45 | 85.15 | 91.93 | 82.12 | 73.54 | 93 |
| Pereira et al. [45] | SWE Pretrained CNN18 | 83 | - | - | - | - | 80 |
| Qin et al. [46] | Pretrained VGG16 Ex-reFus with SPP | 94.7 | 92.77 | 97.96 | - | - | 98.77 |
| Săftoiu et al. [47] | MLP (3-layer) | 89.7 | 91.4 | 87.9 | 88.9 | 90.6 | 95 |
| Săftoiu et al. [48] | MLP (2-layer) | 84.27 | 87.59 | 82.94 | 96.25 | 57.22 | 94 |
| Sun et al. [49] | Hybridized model with voting system (compromise approach) | 86.5 | 82 | 89.7 | - | - | 92.1 |
| Udriștoiu et al. [50] | CNN-LSTM | 98.26 | 98.6 | 97.4 | 98.7 | 97.4 | 98 |
| Zhang et al. [51] | Random forest | 85.7 | 89.1 | 85.3 | - | - | 93.8 |
| Zhao et al. [52] 2020 | Random forest | 86.0 | 86.6 | 85.5 | - | - | 93.4 |
| Zhao et al. [53] 2021 | Machine-learning-assisted approach (B-mode + SWE) using KNN-based bagging model | 93.4 | 93.9 | 93.2 | 86.1 | 97.1 | 95.3 |
| Zhou et al. [54] | RBM + Bayesian (UE) | - | 90.21 | 78.45 | - | - | - |
Acc: accuracy; AUC: area under receiver operator characteristic curve; CNN: convolutional neural network; KNN: k-nearest neighbor; LSTM: long short-term memory; MLP: multilayer perceptron; NPV: negative predictive value; Pc: precision; PPV: positive predictive value; RBM: restricted Boltzmann machine; Rc: recall; Sn: sensitivity; Sp: specificity; SWE: shear-wave elastography. Bold typeface indicates the best performance among the methods.