Fig. 6.

Parameter regression Box-and-Whisker plots. Red boxes represent LSQ derived values; blue boxes represent BSP derived estimates across all measurements (volunteers and repetitions). Left column: BSP prior mean values (μ _p, p = D, F and D ^∗), means/medians of LSQ pixel-wise (LSQ _mean and LSQ _median) and region-of interest (ROI) averaged (${\mathit{LSQ}}_{\mathit{\text{mean}}}^{\mathit{ROI}}$ and ${\mathit{LSQ}}_{\mathit{\text{median}}}^{\mathit{ROI}}$) magnitude derived estimates. Note the logarithmic y-scale of the D ^∗ estimates plot. Middle column: parameter estimate variability is displayed as 3.9 × prior standard deviations from the BSP estimation ($3.9\times {\Sigma }_{\mathit{pp}}^{1/2},p=d,f$and d ^∗), as width of the 95% interval of the LSQ estimates and 3.9 × standard deviation over each region-of-interest of the LSQ estimates (scaled by 3.9 to approximate the 95% interval assuming a Gaussian distribution). Right column: parameter uncertainty displayed as median of estimated standard deviation under the posterior distribution (σ _p, p = d, f and d ^∗)