Advanced Neuroimaging to Unravel Mechanisms of Cerebral Small Vessel Diseases

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


Stroke. 2020;51(1):29-37. 

In This Article

Molecular Neuroimaging in cSVD Research

A detailed review of molecular neuroimaging in vascular cognitive impairment, that included data relevant to the field of cSVD research, was published in Stroke in 2016.[3] For this reason, the fundamentals of molecular neuroimaging will only be briefly discussed, and this section will focus on advances that happened since 2016. The introduction of Pittsburgh Compound B (PiB) as a PET tracer capable of labeling brain amyloid deposits revolutionized dementia research.[45] The initial steps that led to the adoption of amyloid PET imaging by cSVD researchers involved proof-of-concept that PiB also labels vascular amyloid.[46,47] These early studies were not designed to evaluate the value of PiB for CAA diagnosis, and PiB itself was never meant to be a commercially available diagnostic tracer—due to its short half-life (≈20 minutes) requiring an onsite cyclotron and a radiochemistry laboratory with expertise in the synthesis of 11C. Development of 18F amyloid PET tracers, and their Food and Drug Administration-approval as potential tools for Alzheimer disease, made it possible to evaluate them as methods to diagnose CAA that could become widely available. There are some major differences in study designs that aim to validate an amyloid imaging marker for CAA diagnosis versus Alzheimer disease. The studies that validated 18F products for Alzheimer disease diagnosis were conducted in older participants with a relatively short expected survival, in whom the ultimate gold standard was full histopathologic evaluation on autopsy. Many of these patients had dementia, and this approach is totally understandable when the purpose is to compare in vivo amyloid PET imaging results to autopsy data. The situation is very different when the objective is to understand the diagnostic value of an amyloid PET tracer for CAA. This diagnosis is typically made in patients over 55 years of age who had a lobar ICH. In survivors of lobar ICH, the diagnosis of probable CAA is made based on clinical-radiological Boston criteria. Because the goal is to obtain proof-of-concept that the tracer binding mainly represents vascular amyloid, patients with dementia and mild cognitive impairment are ideally excluded, to lower the degree of confounding by parenchymal amyloid plaques. Patients with any CMB in deep locations in addition to cortical ICH/CMBs are likely to have HTN-cSVD and therefore missing such deep CMBs either because of scan parameters or investigator inexperience would also result in erroneous inclusion of non-CAA subjects into the CAA group. Finally, it is well-known that about 15% to 25% of cognitively normal older people have amyloid-positive PET scans, a condition known as presymptomatic Alzheimer disease. Although severe involvement is not common, many patients with CAA have mild degree of parenchymal Alzheimer pathology. In that sense, the enrollment criteria would also make a major difference as cognitively healthy lobar ICH survivors with multiple strictly cortical CMBs probably represent a relatively pure CAA cohort, whereas patients with microbleeds enrolled in memory clinics have probably other confounders. All of these points are important to keep in mind when critically reviewing amyloid PET studies in the field of cSVD and CAA in particular.

The first such study aimed to assess the utility of Florbetapir, a commercially available 18F amyloid tracer, in differentiating CAA from HTN-cSVD.[48] High-resolution susceptibility-weighted imaging was acquired in all 10 patients with CAA and 9 patients with HTN-ICH to minimize the risk of missing CMBs in unexpected locations. Patients diagnosed with probable CAA using Boston criteria had a median of 53 strictly cortical CMBs (interquartile range, 11–134). Patients with probable CAA and HTN-ICH survivors had same mean age (67) and similar risk factor profile. Mini-mental state score was 29 to 30 in all patients, so it was a cognitively normal group. The mean global cortical Florbetapir retention was significantly higher in CAA patients (mean standardized uptake value ratio, SUVR±SD=1.41±0.17) when compared to HTN-ICH (SUVR±SD=1.15±0.08, P<0.001). Using a validated binary method to categorize scans as Florbetapir positive versus negative, all CAA patients had a positive scan but only 1 out of 9 HTN-cSVD patients was amyloid positive (Figure 4). Two investigators that assessed all PET scans, blinded to other imaging/clinical data, had full interrater agreement for amyloid ± status. Patients with CAA also had PiB-PET scans and the global tracer retention correlated very strongly between Florbetapir and PiB (r=0.96, P<0.001). Overall, the study provided Class II evidence that Florbetapir PET provides 100% sensitivity and 89% specificity for determination of probable CAA in cognitively normal patients within the appropriate context.[48] Another study compared a cohort of probable CAA patients diagnosed with modified Boston criteria (only 80% with lobar microbleeds, two-third with cortical superficial siderosis) to HTN-ICH, in the acute phase of ICH without cognitive testing. This study again showed significantly higher mean global SUVR in patients with CAA compared with HTN-ICH (1.27±0.12 versus 1.12±0.12, respectively, P=0.001).[49] Interestingly, the sensitivity (0.6) of the visual assessment was low despite similar specificity (0.89) of Florbetapir for discriminating patients with CAA from HTN-ICH in this study. It will not be possible to rule out the probability that there might have been patients misclassified into the CAA category and that Florbetapir PET revealed more correct results than MRI-based criteria in this study but overall, the current data do not allow making clear generalizations about benefits of amyloid PET imaging for CAA diagnosis. The observed discordance also provides a cautionary tale about problems with in vivo CAA diagnosis in different cohorts using the modified clinical-radiological Boston criteria. A detailed discussion of these issues is provided in the above paragraph. More recent research shed light into some of these problems.

Figure 4.

Illustrative examples of cerebral amyloid angiopathy (CAA) with a positive Florbetapir scan (no contrast between cortex and white matter confirming high cortical amyloid load) and negative scan (clear contrast between cortex and white matter, low cortical tracer uptake) in a patient who had hypertensive deep intracerebral hemorrhage (HTN-ICH). Last pair of scans shows a patient with deep HTN-ICH but positive amyloid scan, a false positive for CAA diagnosis. MRI indicates magnetic resonance imaging; PET, positron emission tomography; and SWI, susceptibility-weighted imaging.

Two very recent studies compared the clinical and imaging characteristics and PiB-PET results of patients with ICH and CMBs in both deep and lobar locations (mixed-location ICH/CMB) separately to strictly lobar ICH/CMB (CAA) and strictly deep ICH/CMB (HTN-cSVD) patients.[50,51] The novel clinical-radiological category, Mixed-location ICH/CMB, makes up 20% to 58% of all primary ICH patients in predominantly white and Southeastern Asian populations, respectively. Clinical, laboratory and MRI features suggested a more severe form of HTN-cSVD as the predominant pathology in patients who have ICH/MBs in both deep and cortical/lobar locations. Amyloid imaging also confirmed significantly higher PiB retention in CAA (mean SUVR=1.43) when compared to both mixed-location ICH/CMBs (mean SUVR=1.06, P=0.003) and deep HTN-ICH/CMBs (mean SUVR=1.1, P=0.002).[51] These recent studies demonstrate the value of using amyloid PET imaging as a surrogate for the CAA pathology. Previous studies showed that CAA-related cortical ICH/CMBs originate from sites with higher baseline vascular amyloid deposits, a finding that was later confirmed in elegantly designed animal studies.[52,53] Human studies using PiB-PET also confirmed probable cause-effect relationships between vascular amyloid load and WMH, lobar lacunes as well as structural network alterations in patients with CAA.[26,39,54] It is probably safe to say that the use of molecular imaging will continue to be an invaluable tool to understand the associations between the molecular changes and disease processes involved in cSVD.