Patients with dementia including Alzheimer’s disease (AD) and vascular dementia (VaD) have both white matter tissue deterioration derived from diffusion tensor imaging (DTI) (1) and increased extracellular water content based on fluid-attenuated inversion-recovery (FLAIR) (2). However, the role of these abnormalities in the etiology of dementia subtypes, particularly with and without cerebrovascular disease (CVD), is not well understood (3). Free-water imaging is a novel analysis method to differentiate the water compartment in the extracellular space from the tissue compartment in a voxel-based manner using DTI data (4). Currently，water and tissue compartment abnormalities in dementia patients especially these with CVD are largely unknown. In addition, the associations between the two compartments and cognitive performance remain unclear.

To fill these gaps, our aims are: 1) to examine the region-specific changes of water and tissue compartments in three dementia subtypes (AD, AD with CVD, and VaD) by applying free-water method on DTI data; 2) to evaluate the spatial similarity between the degree of increased extracellular water derived from FLAIR And DTI data; 3) to examine the associations between water and tissue compartments and symptom severity in the three dementia groups.

T1-weighted structural MRI, FLAIR and DTI of 41 AD, 42 AD with CVD (AD +CVD), 19 VaD, and 60 controls (no cognitive impairment without CVD) was collected at NUS (Siemens, 3Tesla, Tim Trio). White matter hyperintensity (WMH) were obtained for each subject from FLAIR imaging using an in-house automatic segmentation procedure (5). The DTI data were preprocessed by FSL (http://www.fmrib.ox.ac.uk/fsl). Then, we applied tract-based spatial statistics (6) to carry out a voxel-wise analysis of DTI data within major white matter pathways throughout the whole brain. Each participant's aligned fractional anisotropy (FA) data were then projected onto the skeleton, resulting in subject-level skeletonized FA images. Free-water method was applied to differentiate the free-water compartment (FW) and tissue compartment (FA_T) from each subject (Pasternak et al., 2009). Correlations between the WMH percentage and mean free-water value in each white matter tract were calculated. Their associations with dementia severity (Clinical Dementia Rating sum of boxes) were evaluated using a generalized linear model, controlling for age, gender, handedness and ethnicity.

AD, AD+CVD, and VaD had a reduction in FA compared to healthy controls (thresholded at p < 0.005, TFCE corrected). AD+CVD and VaD had more wide spread FA reduction than AD. Moreover, after regressing out the total volume of WMH, the observed FA reductions in AD+CVD and VaD groups were largely decreased (Fig.1). Considering the higher WMH volume in AD+CVD and VaD compared to AD, our findings suggested that WMH might contribute to the deterioration of white matter integrity (i.e. FA reduction).

Compared to the FA values without free-water modeling, there was less deterioration of the tissue compartment (FA_T) in the three dementia groups. In parallel, there was less tissue damage difference between AD+CVD/VaD and AD (mainly cortical region, sparing the subcortical regions), compared to the original FA without free-water modeling (Fig.2). These findings suggested that the original FA might overestimate the tissue damage, part of which might be due to increased extracellular water. Tissue compartment (FA_T) rather than FA might better indicate the actual tissue damage in disease (4).

Moreover, consistent with WMH volume differences, AD, AD+CVD, and VaD had greater free-water than health controls while AD+CVD and VaD had greater free-water than AD. Mean free-water values (averaged across all WM voxels) correlated positively with total WMH volume across all subjects (r=0.58, p<0.01). Our findings suggested that free-water and WMH might reflect a similar pathology, possibly neuroinflammation in dementia.

Lastly, we found that mean free-water value was associated with dementia severity across all dementia patients (r=0.31, p<0.01). No associations with dementia severity were detected for total WMH volume (r=0.11, p=0.28).

With free-water modeling, region-specific and less severe (compared to the tradition method) tissue compartment damage was detected in dementia groups (especially for AD+CVD and VaD). The degree of free-water increase was associated with WMH volume; more importantly, it correlated with dementia severity in patients. Our findings suggested the potential of free-water method in differential diagnosis and disease progression monitoring.