TABLE 3.
Classification performance of deep learning models and radiologists.
| AUC (95% CI) | Accuracy (%) | Sensitivity (% | Specificity (%) | PPV (%) | NPV (%) | |
|---|---|---|---|---|---|---|
| Junior radiologists | 0.740 (0.70–0.75) | 70.6 | 89.3 | 58.7 | 58.0 | 89.5 |
| Senior radiologists | 0.794 (0.72–0.83) | 75.7 | 95.9 | 62.9 | 62.3 | 95.9 |
| B-mode-Net | 0.820 (0.70–0.83) | 74.5 | 75.0 | 77.0 | 73.4 | 62.3 |
| CEUS-mode-Net | 0.815 (0.75–0.89) | 73.9 | 73.8 | 73.2 | 72.5 | 62.2 |
| MUF-Net | 0.877 (0.83–0.93) | 80.0 | 80.4 | 79.1 | 86.9 | 70.0 |
CI, confidence interval; CEUS-mode, contrast-enhanced ultrasound mode; MUF-Net, multimodal ultrasound fusion network; AUC, area under the receiver operating characteristic curve; PPV, positive predictive value; NPV, negative predictive value.