Introduction

White matter hyperintensities (WMH) are one of the commonest features of cerebral small vessel disease, which are visible as hyperintense lesions on T2w imaging.¹ WMH can have serious clinical consequences including decline in cognition² and gait³, and an increase risk of dementia, stroke and death.⁴ Progression of WMH over time is widely examined as it is related to worsening clinical symptoms. Some factors related to WMH progression are higher WMH volume at baseline and older age.⁵ Interestingly, some studies have found regression of WMH, suggesting that WMH are dynamic.^6-9 However, not much is known about WMH regression as it is often discarded as a measurement error. This means that we do not know what causes WMH regression and know very little about other MRI measures associated with this regression. Diffusion MRI is shown to be able to detect underlying structural changes of WMH and normal-appearing white matter (NAWM), even before they are visible on conventional MRI.^{10, 11} We therefore investigate WMH regression and their dynamic properties using diffusion MRI.

Methods

We recruited patients with recent mild lacunar or cortical strokes in an ongoing prospective observational study. After informed consent we acquired clinical, demographic and imaging data within three months after stroke as a baseline measure and a 1 year follow-up visit.¹² Ethical approval was granted by Southeast Scotland Regional Ethics Committee (Ref 18/SS/0044).
At both baseline and the 1-year follow-up, patients underwent MRI scans on a 3T Prisma scanner (Siemens Healthcare, Germany). This included a standard neurovascular imaging protocol (T1w, T2w, FLAIR, and diffusion MRI). We segmented WMH, defined according to STRIVE criteria¹, semi-automatically with a previously mentioned approach and manually delineated stroke masks.¹³ We identified areas of WMH change over time by comparing WMH and NAWM masks between baseline and follow-up visits. This resulted in areas of WMH growth calculated as: areas of positive values when subtracting WMH mask at follow-up from WMH mask at baseline; WMH shrinkage: areas of positive values when subtracting WMH mask at baseline from WMH mask at follow-up; stable WMH: the area of the WMH mask at baseline that intersects with the WMH mask at follow-up; stable NAWM: the area of the NAWM mask at baseline intersecting with the NAWM mask at follow-up. Figure 1 shows an example of these regions in one participant.
Diffusion data were acquired using single-shell diffusion imaging from which we reconstructed diffusion tensor data to obtain fractional anisotropy (FA) and mean diffusivity (MD)¹², and measured them within the white matter areas defined above. We calculated the Peak width of Skeletonized Mean Diffusivity (PSMD) using a previously described protocol,¹⁰ and plotted this with the data divided in groups by quintiles of net WMH volume change (total WMH volume at follow-up minus total WMH volume at baseline).

Results

Preliminary data provides information of 94 participants (32% female) aged 67.5±9.7 years, (range 40.2-85.5) with diffusion MRI data at both baseline and 1 year follow-up. Total WMH volumes at baseline and follow-up were 14.73±17.23 mL and 15.76±19.00 mL respectively. On average WMH volume increased by 1.02 mL (SD=2.80, range -4.28-12.84 mL). With 40% presenting a global decrease.
Comparing areas of white matter changes, baseline and 1-year follow-up FA values (Figure 2) show a similar pattern with highest values for “NAWM stable” followed by “WMH growth”. Areas of “WMH stable” and “WMH shrink” have lowest FA values. MD values (Figure 3) show the opposite pattern, i.e. lowest values for “NAWM stable” and highest values for “WMH stable”.
PSMD results (Figure 4) per quintile of WMH change show higher PSMD for the most increase in WMH volume (Q5), but lowest PSMD for Q3, which has the least net WMH change over time.

Discussion

Areas of NAWM at baseline, whether stable NAWM or developing into WMH, show higher FA and lower MD values than the stable WMH or areas of WMH shrinkage. NAWM at baseline that becomes WMH had lower FA and higher MD at baseline than stable NAWM, indicating some underlying tissue changes that are still invisible. Shrinking WMH had similar FA but lower MD at baseline than stable WMH. PSMD results also suggest that areas of little WMH change (Q3) have better vascular health.
The results suggest MD and PSMD might be more sensitive to initial and subtle structural changes than FA, confirming findings from other studies.^{14, 15}
As this is preliminary data from an ongoing study, it is of descriptive nature and limited. This might have led to a fewer women in this sample, however, women are in general less represented in stroke research.¹⁶ Future work would involve a larger sample size, clinical information including cognitive data, and other imaging measures including T1 mapping, cerebrovascular reactivity and blood-brain barrier permeability to give more insight in the underlying processes.

Conclusion

WMH are dynamic, showing both regression and progression in individuals. Using diffusion tensor imaging, structural changes can be measured, and seem to be present even before these changes are visible on conventional MRI. In particular, MD and PSMD provide more sensitive information on local and global areas of WMH change. Diffusion results suggest that WMH regression is real and should not be regarded as an error or artefact.

References