Table 1.
Quantitative assessment: median (inter-quartile range) Dice similarity coefficient (DSC), Hausdorff distance (HD) and proportion (95% CI) of correctly labeled vertebrae. The sequential approach outperformed the combined approach during the external validation.
| Sequential approach | Combined approach | p | |
|---|---|---|---|
| Internal validation | |||
| DSC (%) | |||
| Thoracic vertebrae | 96.1 (95.0–96.9) | 96.2 (95.0–97.0) | 0.003 |
| Lumbar vertebrae | 97.6 (97.3–97.9) | 97.6 (97.1–97.9) | 0.004 |
| All vertebrae | 96.7 (95.5–97.4) | 96.7 (95.4–97.4) | 0.13 |
| HD (mm) | |||
| Thoracic vertebrae | 3.7 (2.8–5.4) | 4.1 (2.8–5.8) | 0.45 |
| Lumbar vertebrae | 3.2 (2.4–4.0) | 3.0 (2.2–4.2) | 0.72 |
| All vertebrae | 3.6 (2.8–5.1) | 3.6 (2.4–5.7) | 0.66 |
| Labeling (%) | |||
| All vertebrae | 90.7 (88.3–93.1) | 91.6 (89.3–93.8) | |
| External validation | |||
| DSC (%) | |||
| Thoracic vertebrae | 93.4 (90.7–94.8) | 93.9 (88.9–95.1) | 0.09 |
| Lumbar vertebrae | 96.0 (95.0–96.4) | 95.4 (92.8–96.1) | <0.001 |
| All vertebrae | 94.5 (91.8–95.8) | 94.4 (91.4–95.5) | <0.001 |
| HD (mm) | |||
| Thoracic vertebrae | 4.6 (3.6–5.9) | 7.1 (3.6–21.3) | <0.001 |
| Lumbar vertebrae | 4.0 (3.2–6.1) | 7.2 (4.0–12.4) | <0.001 |
| All vertebrae | 4.5 (3.4–6.0) | 7.1 (3.7–15.1) | <0.001 |
| Labeling (%) | |||
| All vertebrae | 79.6 (74.0–85.2) | 55.7 (48.9–62.6) | |