Table 1. Properties of training and testing data.
| Characteristic | Training set (training: validation = 9:1 randomly distributed) | Positive test set | Negative test set |
|---|---|---|---|
| Data source | Hospital 1 | Hospital 2 | Hospital 2 |
| Number of patients | 394 | 214 | 123 |
| Male/ Female | 303/ 91 | 154/ 60 | 90/ 33 |
| Age (mean, SD) | (64.9, 11.1) | (62.6, 11.2) | (59.9. 9.7) |
| Dynamic/ Non-dynamic CT | 376/ 18 | 208/ 6 | 117/ 6 |
| Number of liver CT images | 10130 | 1833 | 20283 |
| Number of tumor ROIs | 11258 | 1874 | - |
| Number of hyperdense ROIs | 3078 | 334 | - |
| Number of hypodense ROIs | 6544 | 1313 | - |
| Number of heterogeneous density ROIs | 1636 | 227 | - |
| LD of GT tumor ROI (mean, SD) (cm) | (6.37, 3.97) | (4.75, 2.16) | - |
| Area of GT tumor ROI (mean, SD) (cm 2 ) | (32.08, 37.29) | (17.03, 15.83) | - |
LD: longest diameter.
ROI: region of interest.
GT: ground truth.
SD: standard deviation.