Table 3.
Biomarker | Bone | Method | Variable | Estimate | Standard error | p-value | ROC (sensitivity/specificity/AUC) | Likelihood ratio p-value |
---|---|---|---|---|---|---|---|---|
Cartilage thickness (n = 2151) | Femur | Classical: clinical compartment average | Intercept | − 3.59 | 0.621 | 7.56 × 10–9 |
13.5 95.2 63.3 |
7.01 × 10–33 |
Age | − 0.011 | 5.1 × 10–3 | 3.06 × 10–2 | |||||
BMI | 0.077 | 1.01 × 10–2 | 1.84 × 10–14 | |||||
Sex | − 0.193 | 0.114 | 9.05 × 10–2 | |||||
LF thickness | − 0.582 | 0.264 | 2.77 × 10–2 | |||||
MF thickness | 1.289 | 0.246 | 1.77 × 10–7 | |||||
Proposed: DL-guided weighted average | Intercept | − 2.52 | 0.645 | 9.16 × 10–5 |
33.5 89.8 69.2 |
|||
Age | − 1.97 × 10–2 | 5.36 × 10–3 | 2.33 × 10–4 | |||||
BMI | 5.14 × 10–2 | 1.06 × 10–2 | 1.13 × 10–6 | |||||
Sex | − 8.78 × 10–2 | 0.118 | 0.455 | |||||
LF thickness | 2.25 | 0.364 | 6.55 × 10–10 | |||||
MF thickness | 2.29 | 0.268 | 1.57 × 10–17 | |||||
DL-thickness | − 3.66 | 0.32 | 2.02 × 10–30 | |||||
Tibia | Classical: clinical compartment average | Intercept | 0.496 | 0.638 | 0.437 |
15.5 94.2 63.6 |
1.93 × 10–14 | |
Age | − 1.81 × 10–2 | 5.17 × 10–3 | 4.5 × 10–4 | |||||
BMI | 0.078 | 0.01 | 7.27 × 10–15 | |||||
Sex | − 0.356 | 0.106 | 8.0 × 10–4 | |||||
LT thickness | − 0.445 | 0.182 | 1.47 × 10–2 | |||||
MT thickness | − 0.537 | 0.15 | 3.42 × 10–4 | |||||
Proposed: DL-guided weighted average | Intercept | 0.81 | 0.65 | 0.213 |
24.8 92.9 66.9 |
|||
Age | − 2.11 × 10–2 | 5.27 × 10–3 | 6.06 × 10–5 | |||||
BMI | 7.03 × 10–2 | 1.02 × 10–2 | 4.86 × 10–12 | |||||
Sex | − 0.387 | 0.107 | 3.12 × 10–4 | |||||
LT thickness | 0.289 | 0.208 | 0.165 | |||||
MT thickness | 0.108 | 0.173 | 0.533 | |||||
DL-thickness | − 1.37 | 0.184 | 9.6 × 10–14 | |||||
Patella | Classical: clinical compartment average | Intercept | 1.21 | 0.644 | 6.05 × 10–2 |
15.8 94.8 65.0 |
0.851 | |
Age | − 2.38 × 10–2 | 5.33 × 10–3 | 8.09 × 10–6 | |||||
BMI | 6.64 × 10–2 | 1.03 × 10–2 | 1.06 × 10–10 | |||||
Sex | − 0.389 | 0.105 | 2.28 × 10–4 | |||||
L thickness | − 0.398 | 0.118 | 7.12 × 10–4 | |||||
M thickness | − 0.424 | 0.12 | 3.97 × 10–4 | |||||
Proposed: DL-guided weighted average | Intercept | 1.215 | 0.645 | 5.95 × 10–2 |
16.0 94.9 65.0 |
|||
Age | − 2.38 × 10–2 | 5.33 × 10–3 | 7.96 × 10–6 | |||||
BMI | 6.63 × 10–2 | 1.03 × 10–2 | 1.20 × 10–10 | |||||
Sex | − 0.39 | 0.106 | 2.24 × 10–4 | |||||
L thickness | − 0.376 | 0.166 | 2.39 × 10–2 | |||||
M thickness | − 0.401 | 0.173 | 2.06 × 10–2 | |||||
DL-thickness | − 4.57 × 10–2 | 0.243 | 0.851 |
The demographic factors, such as age, BMI, and sex, are included to the logistic regression models as well as the different cartilage thickness averaging methods. The results are shown for the two definitions for OA imaging biomarkers, clinical compartment average and DL-guided weighted average for the femur, tibia, and patella.
LF Lateral Femur, MF Medial Femur, MT Medial Tibia, LT Lateral Tibia, M Medial, L Lateral.