Gray matter contamination in arterial spin labeling white matter perfusion measurements in patients with dementia

Abstract

Introduction: White matter (WM) perfusion measurements with arterial spin labeling can be severely contaminated by gray matter (GM) perfusion signal, especially in the elderly. The current study investigates the spatial extent of GM contamination by comparing perfusion signal measured in the WM with signal measured outside the brain.

Material and methods: Four minute 3T pseudo-continuous arterial spin labeling scans were performed in 41 elderly subjects with cognitive impairment. Outward and inward geodesic distance maps were created, based on dilations and erosions of GM and WM masks. For all outward and inward geodesic distances, the mean CBF was calculated and compared.

Results: GM contamination was mainly found in the first 3 subcortical WM voxels and had only minor influence on the deep WM signal (distances 4 to 7 voxels). Perfusion signal in the WM was significantly higher than perfusion signal outside the brain, indicating the presence of WM signal.

Conclusion: These findings indicate that WM perfusion signal can be measured unaffected by GM contamination in elderly patients with cognitive impairment. GM contamination can be avoided by the erosion of WM masks, removing subcortical WM voxels from the analysis. These results should be taken into account when exploring the use of WM perfusion as micro-vascular biomarker.

Keywords: ASL, arterial spin labeling; Arterial spin labeling; CBF, cerebral blood flow; CSF, cerebrospinal fluid; Dementia; GM, gray matter; Gray matter contamination; PSF, point spread function; PV, partial volume; Partial volume; SNR, signal-to-noise ratio; WM, white matter; White matter perfusion.

Fig. 1

Single slice distance analysis pipeline visualized for a single patient: tissue probability maps (a, A) converted into masks (b, B), gaps filled (c, C), erosion and dilation (d, D) and the resulting city-block geodesic maps (e, E). Lower and upper cases represent WM and GM respectively. In the right lower corner the ASL slice is shown for reference.

Fig. 2

a–c show single slice distance analysis (left column) and partial volume analysis (right column). a) mean CBF; b) mean GM-WM CBF ratio; c) mean number of voxels. The distance numbers on the left x-axis correspond with distances in Fig. 1. Distances − 1 to − 7 represent GM mask dilation steps, distances 1 to 7 represent WM mask erosion steps. Distance 0 represents the mean GM CBF (tissue probabilities > 90%). The numbers on the right x-axis represent bins of the WM tissue probabilities (bin size 1%). Lines and planes represent mean values and ± 95% CI respectively. Significant differences (p < 0.001) between negative and positive distances are indicated by an asterisk (*).

Fig. 3

a–c visualize the difference between two masks obtained by either a) the exclusion of partial volume (PV) voxels by thresholding the WM mask at a tissue probability of 100% or b) the application of three erosions on a large WM mask (tissue probabilities > 10%). The same WM distance map color scale (Fig. 1c, reprinted here for reference (c)) is applied here to visualize the position of the voxels included in both masks.