Advanced Neuroimaging to Unravel Mechanisms of Cerebral Small Vessel Diseases

M. Edip Gurol, MD, MSc; Geert J. Biessels, MD; Jonathan R. Polimeni, PhD

Disclosures

Stroke. 2020;51(1):29-37. 

In This Article

Advances in Structural Neuroimaging of Cerebral Microangiopathies

Types of cSVD

Multiple lines of evidence show that cSVDs result in both ischemic and hemorrhagic brain lesions. The 2 most common causes are hypertensive arteriolosclerosis and cerebral amyloid angiopathy (CAA). Hypertension is an exceedingly prevalent condition found in 63.1% of adults over 60 years of age in the United States, and hypertensive cSVD (HTN-cSVD) is the most common type of cSVD.[8,9] CAA refers to the accumulation of Aβ (amyloid-β peptides) in the walls of the leptomeningeal and cortical vessels.[10] Lacunar infarcts, white matter hyperintensities (WMH), ICH, and cerebral microbleeds (CMB) are tissue lesions detectable with brain imaging that are manifestations of cSVDs.[4] Neuroimaging studies have shown that topographical distribution of these lesions correlates with the type of the underlying microangiopathy. Recent work using both standard and advanced structural magnetic resonance imaging (MRI) also added new lesion types to the list of the classically accepted cSVD-related pathologies.

ICH and CMBs in deep-brain regions (basal ganglia, thalamus, and pons) are associated with severe HTN-cSVD, whereas ICH/MBs limited to superficial/lobar brain regions and cortical superficial siderosis are mainly seen in patients with CAA.[3] The presence and distribution of these hemorrhagic lesions still constitute the most important markers in the diagnosis and etiologic classification of cSVD in living individuals. Besides its diagnostic relevance, cortical superficial siderosis is also a very important marker of ICH recurrence risk in patients with CAA.[11,12] Use of susceptibility-weighted imaging MR at ultrahigh magnetic field strength (7 Tesla) answered an important question in CAA research, demonstrating that CAA-related microbleeds exclusively occur within the cortex (Figure 1).[13] Furthermore, the use of 7T susceptibility-weighted imaging MR revealed high incidence of novel hemorrhagic markers in hereditary cerebral hemorrhage with amyloidosis-Dutch type, namely intragyral hemorrhage (47% versus 0% in controls, P<0.001), and a high incidence of striped cortex sign (40% versus 0% in controls, P<0.005).[14] More widespread use of 7T susceptibility-weighted imaging in research is likely to result into better understanding of hemorrhagic lesions seen within the context of cSVDs.

Figure 1.

Impact of scanner/parameters on microbleed detection. Optimized susceptibility-weighted imaging obtained in Ultra-High Field (7T) magnetic resonance imaging (MRI) scanner (A) shows a much higher number of cortical microbleeds when compared with clinical-grade MRI (B) obtained on the same day in a patient with cerebral amyloid angiopathy.

Ischemic Markers of cSVD

White matter disease, also called leukoaraiosis, is defined as hypodensities on computerized tomography and WMH on T2/fluid attenuated inversion recovery MRIs in subcortical and periventricular regions that do not show cavitation or other characteristics of a completed infarction.[15] Multiple lines of evidence established leukoaraiosis as an important marker of cSVD. WMH volume was found to be significantly higher in patients with CAA and HTN-cSVD when compared with age-similar healthy adults as well as patients with mild cognitive impairment and Alzheimer disease.[16–18] WMH volume also correlated strongly with markers of severity of cSVD, both cross sectionally and longitudinally.[19,20] There has been remarkable interest in identifying WMH patterns that could distinguish different types of cSVD and even cSVD from other pathologies such as Alzheimer disease. Studies that used simple voxel-based comparisons failed to identify significant differences in WMH distribution probably because of the fact that the bulk of WMH is periventricular in all patients, therefore, no clear patterns emerged across different diagnostic categories after adjusting for age and the volume of leukoaraiosis.[17,18] A recent study that used visually identified WMH patterns showed differences between patients with CAA and HTN-cSVD. The presence of peribasal ganglia linear WMH was more commonly found in HTN-cSVD when compared with CAA (19% versus 7.8%; P=0.001) whereas having 10 or more small circular WMHs (subcortical spots) outside of periventricular locations was associated with the diagnosis of CAA (29.8% versus 16.8%; P=0.004).[21] One retrospective study that evaluated the center of WMH over the anteroposterior axis using computer-assisted segmentations suggested a more posterior distribution in patients with pathological evidence of CAA but no ICH.[22] More recent studies use deep learning to create segmentation algorithms and perform voxel-based spectral clustering analysis on aligned WMH maps to group image voxels into clusters, maximizing within-group and minimizing between-group similarities.[23] These studies confirm the previous findings of a deep WMH distribution in HTN-cSVD versus more peripheral and posterior WMH patterns in CAA.[23,24]

Classical lacunar infarcts are found distal to occlusive lesions of small perforating arteries mostly in basal ganglia, internal capsule, thalami, and pons—regions known to be supplied by deep branching/perforating vessels. These lesions are caused by distinct vessel pathologies that are categorized based on the size of the vessel: arteriosclerosis/atherosclerosis, arteriolosclerosis, and lipohyalinosis, collectively categorized as HTN-cSVD. There has been growing interest in lacunar infarcts in primary cSVD-related ICH as it is possible to identify the predominant microangiopathy in these patients during life. Lacunes were found in 25% of primary ICH patients. Lacunes in deep-brain locations were more commonly seen in patients with hypertensive deep ICH (15.2%) when compared to patients with CAA (2.1%, P<0.001). Interestingly, radiological lacunes in more superficial subcortical locations (lobar lacunes) were more common in CAA-ICH (20.4%) than HTN-ICH (5.7%, P<0.001; Figure 2).[25] Another study also found more common presence of lobar lacunes in CAA-ICH than in HTN-ICH (29.2% versus 11.6%, P=0.036), as well as a significant correlation between lobar lacune counts and brain amyloid load (r=0.40, P=0.02).[26] These data establish radiological lacunes in lobar (nondeep) locations to be associated with CAA, a finding that might have both diagnostic and therapeutic implications.

Figure 2.

Distribution of lacunes in cerebral microangiopathies. Examples of lobar lacune (centrum semiovale) in cerebral amyloid angiopathy (CAA) (A), deep lacune in hypertensive deep intracerebral hemorrhage (HTN-ICH) (B) and topographical distribution maps of lacunes in these brain microangiopathies (C). Reproduced from Pasi et al25 with permission. Copyright ©2017, Wolters Kluwer Health, Inc.

Cerebral microinfarcts have classically been a histopathologic construct representing microscopic lesions with a mean diameter between 0.2 and 1.0 mm, found at autopsy in 16% to 46% of unselected elderly people dying of all causes.[27] A study that identified punctate diffusion-weighted imaging restricted lesions in 15% of patients who had a lobar ICH, but none in patients diagnosed with Alzheimer disease, was the first that suggested the possibility to detect microinfarcts in the setting of severe cSVD.[28] Another study identified similar rates of microinfarct detection but somehow different distributions on diffusion-weighted imaging in HTN-cSVD and CAA patients both during the acute phase of ICH and in long-term follow-up.[29] Another study that used longitudinal diffusion tensor imaging demonstrated that asymptomatic diffusion-weighted imaging lesions produce chronic local microstructural injury in patients with CAA.[30] The major pitfall of diffusion-weighted imaging to detect microinfarcts is that the lesions are only visible during the acute/subacute phases of the infarct, that is, for 7 to 14 days. The use of ultrahigh-field strength (7T) and high-resolution imaging allow the detection of cortical microinfarcts beyond the acute phase on T1, T2, and fluid attenuated inversion recovery MRI but the vast majority of microinfarcts still remain under the detection limits of clinical in vivo MRI.[31] The presence of cerebral microinfarcts on 3T MRIs has been associated with vascular cognitive impairment and dementia in ischemic stroke and memory clinic cohorts.[32] It should also be remembered that both potential sources of cardiac embolism and other proximal embolic sources were associated with cerebral microinfarcts, so the ideal research in this field should account for these potential etiologies before attributing all microinfarcts to cSVD.[33,34]

Markers of Global Structural Injury Related to cSVD

Cortical atrophy is an important contributor to cognitive impairment and dementia in older adults. One of the major setbacks to determine the association between cSVD and cortical atrophy have been the common co-occurrence of parenchymal Alzheimer pathology such as senile plaques and neurofibrillary tangles in patients who have brain microangiopathies, such as CAA. A recent study compared the cerebral cortical thickness between patients with hereditary cerebral hemorrhage with amyloidosis-Dutch type (HCHWA-D, a hereditary type of relatively pure CAA, without parenchymal Alzheimer pathology) and age-matched healthy controls (mean age 46.7). HCHWA-D had significantly thinner cortex when compared to healthy controls (2.31 versus 2.42 mm, P=0.006; Figure 3).[35] Similar findings of cortical atrophy were confirmed in sporadic CAA (mean age=72 and mean cortical thickness=2.17 mm) when compared to 2 separate age-matched healthy control groups (2.31 and 2.27 mm). Patients with Alzheimer disease had more pronounced cortical atrophy when compared to age-matched sporadic CAA. A physiological measure of vascular dysfunction obtained using functional MRI (fMRI) methods accounted for 63% of the total effect of CAA on cortical atrophy, using mediation analyses. These fMRI measures will be further discussed under the physiological imaging section below. This study, overall, is a good example of the benefits of using advanced anatomic and physiological imaging modalities in appropriate cohorts to tackle important questions that cannot be answered with simpler designs. There is indirect evidence suggesting that hypertension might affect cortical atrophy as well, however, studies that directly test these associations in well-characterized HTN-cSVD cohorts are needed.[36,37]

Figure 3.

Topographical surface maps showing regions with significant cortical thinning in patients with a hereditary form of cerebral amyloid angiopathy (CAA) when compared with age-matched healthy controls. Reproduced from Fotiadis et al35 with permission. Copyright ©2016, Elsevier.

Disruption of brain networks is a postulated mechanism through which cSVD can result in loss of brain function. Diffusion tensor imaging is another advanced imaging modality that provides insights into microstructural changes in brain tissue and cerebral network connectivity. Early work in the field of cSVD showed a pattern of regional brain tissue degeneration in the temporal lobe white matter and the splenium of the corpus callosum in CAA.[38] A recent advance was the use of graph theory to characterize the efficiency of the brain networks, enabling the researchers to calculate measures of global network efficiency. The use of these techniques together with amyloid positron emission tomography (PET) imaging and detailed cognitive testing showed that reduced structural brain network efficiency might mediate the relationship between advanced CAA and neurological dysfunction.[39] One potential weakness of diffusion tensor imaging is the relatively wide variability in scan parameters and the variability introduced by different analytic techniques that limit the use of this otherwise important MRI method. Recent efforts focused on developing a robust diffusion tensor imaging MRI marker based on skeletonization of white matter tracts and histogram analysis. The resultant fully automated marker, peak width of skeletonized mean diffusivity, showed excellent correlation with processing speed outperforming conventional markers (WMH, lacune, and brain volumes) in different populations, such as hereditary and sporadic cSVDs and healthy controls.[40] If these results are reproduced in the hands of independent investigators, peak width of skeletonized mean diffusivity can become an important tool in cSVD-related vascular cognitive impairment research.

Perivascular spaces (PVS or Virchow-Robin spaces) have been noted on brain imaging since the initial use of MRIs in 1980 to 1990s. Over the past years, there has been increasing number of studies that identified high numbers of enlarged PVS in cSVD patients. Specifically, high counts of enlarged PVS in centrum semiovale were more common in CAA, whereas more basal ganglia enlarged PVS were found in HTN-cSVD.[41,42] Recent studies have shown that the use of optimized T2-weighted pulse sequences with 7T scanners significantly improves the detection of PVS. Using such high-sensitivity scans, PVS counts were not associated with age or vascular risk factors but related to cSVD markers such as CMB counts and WMH volumes.[43] Another 7T MRI study showed that enlarged juxtacortical PVS colocalize with cortical CMBs in patients with mild cognitive impairment and Alzheimer disease, suggesting a common underlying cause, probably CAA.[44] Overall, it is likely that research into PVS will increasingly use high-resolution MRI, allowing the investigators to better understand the real meaning of these imaging markers. There are certainly other promising advanced structural neuroimaging approaches that can be highly relevant for clinical research into mechanisms of cSVD. Such methods include advanced connectome imaging modalities, quantitative susceptibility mapping, and others. They will not be reviewed as their application to cSVD-related research has not yet been established.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....