Table 2 Summary of the performance metrics of the prediction models for lumbar BMD.
From: A prediction model of pediatric bone density from plain spine radiographs using deep learning
Vertebrae | Overall vertebrae | ||||
|---|---|---|---|---|---|
L1 | L2 | L3 | L4 | ||
Number of ROIs (n, [%]) | 60 (100) | 60 (100) | 60 (100) | 58 (96.7) | 238 (99.2) |
Predicted mean BMD (SD, g/cm2) | 0.67 (0.11) | 0.68 (0.11) | 0.69 (0.12) | 0.70 (0.12) | 0.69 (0.12) |
Measured mean BMD (SD, g/cm2) | 0.65 (0.17) | 0.70 (0.19) | 0.72 (0.20) | 0.72 (0.20) | 0.70 (0.19) |
Correlation coefficient (r) | 0.64 | 0.70 | 0.73 | 0.78 | 0.72 |
R square, RMSE | 0.41, 0.13 | 0.48, 0.14 | 0.53, 0.13 | 0.60, 0.13 | 0.51, 0.13 |
MAE | 0.10 | 0.10 | 0.10 | 0.09 | 0.10 |
Calibration slope, | 0.97 | 1.26 | 1.16 | 1.28 | 1.18 |
Bland–Altman bias (g/cm2; SD) | –0.02, 0.13 | –0.02, 0.14 | 0.02, 0.14 | 0.02, 0.13 | 0.01, 0.14 |
