Research Article
Improving Accuracy of Lung Nodule Classification Using Deep Learning with Focal Loss
Table 4
Classification results of LdcNet compared with other works.
| Work | # scans | Dataset | Consensus level | Performance (%) | Accuracy | Sensitivity | Specificity |
| Proposed LdcNet-FL | 888 | LIDC/IDRI | ≥3 | 97.2 | 96.0 | 97.3 | Proposed LdcNet-CE | 888 | LIDC/IDRI | ≥3 | 95.6 | 90.2 | 96.0 | Li et al. [10] | 1,010 | LIDC/IDRI | ≥1 | 86.4 | 87.1 | — | Kuruvilla and Gunavathi [11] | 155 | LIDC/IDRI | ≥2 | 93.3 | 91.4 | 100 | Choi and Choi [12] | 58 | LIDC/IDRI | ≥1 | 97.6 | 95.2 | 96.2 |
|
|