Table 2.
Multivariate analysis assessment of the relationship between spectral data in the different optical windows and subchondral bone parameters.
| N | R2 | RMSECV (%) | RMSEC (%) | |
|---|---|---|---|---|
| Optical window 1* | ||||
| Sb.Th | 6 | 92.3 | 20.5 | 7.1 |
| Tb.Th | 6 | 88.4 | 16.4 | 6.7 |
| BV/TV | 6 | 83 | 25.2 | 9.8 |
| SMI | 5 | 79.7 | 22.5 | 10.8 |
| Optical window 2 | ||||
| Sb.Th | 6 | 43.0 | 25.4 | 19.4 |
| Tb.Th | 5 | 52.3 | 17.6 | 13.7 |
| BV/TV | 4 | 30.6 | 27.6 | 22.9 |
| SMI | 7 | 43.2 | 24.3 | 18.1 |
| Optical window 3 | ||||
| Sb.Th | 6 | 45.9 | 24.4 | 18.9 |
| Tb.Th | 5 | 48.9 | 18.8 | 14.2 |
| BV/TV | 5 | 65.7 | 28.4 | 20.0 |
| SMI | 7 | 47.7 | 24.5 | 17.4 |
[N = number of PLS components; R2 = coefficient of determination; RMSECV = root mean square error of cross-validation; RMSEC = root mean square error of calibration. Both error parameters are estimated relative to the range of the reference data]. R2 for BV/TV was calculated using non-parametric (Spearman) method because of the lack of normality. *Optimal tissue optical window.