Table 3.
Liver fibrosis stage classification performance on the amplitude spectrum test set in Group B by the proposed method, compared to that by the 1D CNN models of Han et al. [7] and Nguyen et al. [8]. The largest value of each metric for each classification is indicated as bold numbers. ACC: accuracy; SEN: sensitivity; SPE: specificity; AUC: area under the receiver operating characteristic curve.
| Fibrosis Stage | Model | ACC (%) | SEN (%) | SPE (%) | AUC |
|---|---|---|---|---|---|
| Ours | 90.48 | 95.24 | 85.71 | 0.941 | |
| ≥F1 | Han et al. [7] | 88.10 | 85.71 | 90.48 | 0.937 |
| Nguyen et al. [8] | 80.95 | 85.71 | 76.19 | 0.853 | |
| Ours | 81.25 | 75.00 | 87.50 | 0.771 | |
| ≥F2 | Han et al. [7] | 78.13 | 87.50 | 68.75 | 0.763 |
| Nguyen et al. [8] | 75.00 | 78.13 | 71.88 | 0.767 | |
| Ours | 74.29 | 63.64 | 92.31 | 0.649 | |
| ≥F3 | Han et al. [7] | 71.43 | 63.64 | 84.62 | 0.621 |
| Nguyen et al. [8] | 71.43 | 68.18 | 76.92 | 0.678 | |
| Ours | 80.56 | 77.78 | 83.33 | 0.758 | |
| ≥F4 | Han et al. [7] | 75.00 | 77.78 | 72.22 | 0.753 |
| Nguyen et al. [8] | 72.22 | 77.78 | 66.67 | 0.691 |