Lacunar Infarcts, but Not Perivascular Spaces, Are Predictors of Cognitive Decline in Cerebral Small-Vessel Disease

Supplemental Digital Content is available in the text.


Image processing
The raw DICOMS were imported using the SPM8 software package (http://www.fil.ion.ucl.ac.uk/spm/software/spm8/). T1-weighted and FLAIR images were coregistered together using an affine transformation in SPM for each individual, before affine transformation to the same orientation as the MNI template and resliced to 1 mm isotropic resolution using 4th degree b-spline interpolation. The conventional SPM segmentation pipelines were adapted and optimised to our study population. This better captured population specific features (such as larger ventricles). To achieve this, a group average template was generated and all the T1-weighted and FLAIR images are warped to this space. The T1weighted and FLAIR images in the group average space were then used to create population specific tissue probability maps (TPMs). The default SPM TPMs were replaced by the population specific TPMs 1 in the SPM segmentation algorithm to re-segment native space images to generate grey matter (GM), white matter (WM), cerebrospinal fluid (CSF) and WMH tissue class segmentation images. These were then combined with manually defined lacune ROIs to provide five tissue class images per individual ( Figure 1 in main manuscript). The WMH segmentation maps were binarised at a threshold set for each individual by checking results manually to ensure accurate correspondence with lesions on the FLAIR image. Results were manually refined where necessary to optimise accuracy. For further details on image processing please refer to Lambert et al 1 .

WMH volume
WMH masks were created using the above technique from the FLAIR images. The volumes at each time-point were calculated in native subject space by summing the binarised corrected segmentations. To adjust for the effects of head size and brain atrophy on WMH volume, we analysed WMH load by expressing WMH volume as a percentage of the volume of total white matter.

Estimation of brain volume
Brain volume at baseline was calculated using a fully automated program, SIENAX (Crosssectional Structural Image Evaluation using Normalisation of Atrophy www.fmrib.ox.ac.uk/fsl) 2 on T1-weighted images.

Cerebral Microbleeds
Baseline CMBs were identified by a consultant neuroradiologist. The Brain Observer Microbleed Rating Scale (BOMBS) was used to identify and describe CMB locations. CMBs were defined as homogeneous round focal areas < 10 mm in diameter of low signal intensity on T2*-weighted GRE images. Only CMBs meeting the BOMBS "certain" criteria were analysed. Bilateral basal calcification, flow voids from blood vessels, and low signals averaging from adjacent bone were regarded as "uncertain" CMBs.

Neuropsychological assessment
Cognitive assessment was carried out annually using well established standardized tests to include measures sensitive to the pattern of cognitive impairment associated with SVD. Premorbid IQ was estimated using the National Adult Reading Test-Restandardised (NART-R) 3 . Tasks were grouped into broad cognitive functions as follows: Executive function: Trail Making test, (Reitan 1996) Modified Wisconsin Card Sorting Test 4 and Phonemic Fluency (FAS) 5 .
Processing speed: Wechsler Adult Intelligence Scale-III (Wechsler, 1997a) Digit symbol substitution 6 , Speed of Information Processing Task 7 and Grooved Pegboard Task 8 .
Task performance was age scaled using published normative data, transformed into z-scores and aggregated to construct the cognitive indices of executive function and processing speed by averaging across the component test measures for each subject, with a Global Function index as an amalgam of all measures. For further details on the cognitive assessment please refer to Lawrence et al 9 .
Supplementary Table I: Estimates (confidence intervals) of the intercept and slope of cognitive indices over 5-year follow-up followed by the estimated effects (confidence intervals) of baseline lacunes and perivascular spaces on these values. Estimates with a significant effect are shown in bold. Lacunes number and volume had a significant effect on all cognitive indices at baseline. Lacune number explained some variability in the slope of executive function and global function in the model. PvS did not have a significant effect on the intercept or slope. Results were considered statistically significant when confidence intervals excluded zero.