the MRI Project Data: Examples of Region- Specific Models
The joint model specified in (17.1) was applied for each combination of the seven MRI parameters and the four histology stains in each of the 23 ROIs (shown in Figure 17.15). In this section, we discuss the results in the motor cortex (Section 17.5.1) and the caudate-putamen (Section 17.5.2) regions. Note that, as explained in Section 17.4, the joint model (17.1) is formulated with a constant effect of histology in the wildtype mice (pYi = ).
The Motor Cortex: GFAP Staining and MRI-AK
The observed data for GFAP percentage of area stained and MRI-AK are shown in Figure 17.9a where an age-dependent shift of MRI and histology measurements for older transgenic mice is observed. Parameter estimates obtained for the joint model are presented in Table 17.2. The estimated regression model, shown in Figure 17.9b is given by вг = 0.064 + 192.675аг. The surrogacy measure at disease level R2D = 0.91 indicates that MRI-AK is a good predictive biomarker for GFAP staining. At an individual level, after adjusting for the disease effect, there is low correlation between the residuals (Pa = Pw = 0.13) indicating a low individual-level surrogacy (Figure 17.10a and b).
Figure 17.11a presents the disease-level surrogacy measures for all seven MRI parameters and four histology stains in the motor cortex region. Note that the MRI-AK parameter is found to be predictive for 4G8 staining (RD = 0.94) and GFAP staining (RD = 0.91), while it has relatively low predictive value for the IBA-1 (RD = 0.60) and MBP (RD = 0.47) staining. MRI-MD was found to be predictive at the disease-level for 4G8 (R2D = 0.87), GFAP (RD = 0.90), and IBA-1 (R2D = 0.76) stainings, respectively. In addition,
The motor cortex region: Evaluation of surrogacy for MRI-AK and GFAP percentage of area stained. Gray symbols: wildtype mice. Black symbols: transgenic mice.
MRI-RD was predictive for MBP staining with R2D = 0.83. The results for individual-level surrogacy are shown in Figure 17.11, which reveals that MRI parameters were not predictive for histology at an individual-level in both transgenic (Figure 17.11b) and wildtype mice (Figure 17.11c).
The motor cortex region: Evaluation of surrogacy for MRI-AK and GFAP percentage of area stained. Residuals.