Repeatability of diffusion-based stiffness prediction - A healthy volunteer study

The study investigated the repeatability of brain diffusion-based stiffness prediction (DWI_stiff) in healthy volunteers.

Thirty-one healthy volunteers were examined with DWI_stiff using two different sets of b-values: b200-1500 s/mm² (DWI_stiff, 1500) and b200-1000 s/mm² (DWI_stiff, 1000). Each b-value set was scanned twice per imaging session without repositioning the participants. DWI_stiff images were reconstructed from each set. Two observers delineated regions of interest (ROIs) on each DWI_stiff image. The repeatability coefficient (RC), coefficient of variation (CV), inter- and intraobserver agreement were calculated.

After excluding three participants due to image artifacts, the study included twenty-eight volunteers (mean age (range)) 37 years (24-62), 10 males, 18 females). For DWI_stiff, 1500, the lowest and the highest RCs were in the parietal lobe (0.52) and respectively the brain stem (1.17). The lowest RC for DWI_stiff, 1000 was in the frontal lobe (0.42) and the highest in the brain stem (1.58). The CV for whole brain measurements was 3.83 % for DWI_stiff, 1500 and 4.93 % for DWI_stiff, 1000. The Bland‒Altman (BA) limits of agreement (LoA) for the intraobserver agreement of DWI_stiff, 1500 were -0.90 to 1.06 and respectively -0.78 to 0.88 for DWI_stiff, 1000. Regarding interobserver agreement, the LoA were -0.85 to 0.94 for DWI_stiff, 1500 and -0.61 to 0.66 for DWI_stiff, 1000.

DWI_stiff is a precise technique with some observer dependence. Repeatability is higher for DWI_stiff, 1000 s/mm² than for DWI_stiff 1500 s/mm².

Our findings suggest that DWI_stiff can reliably detect stiffness changes larger than 4.93 % in healthy volunteers. Further studies should investigate whether the repeatability of DWI_stiff may be affected by the presence of pathology such as a brain tumor.