Clustering of Cerebral Microbleeds in Cerebral Amyloid Angiopathy: A Flow Territory-Based Analysis

Abstract

Background

Cerebral microbleeds (CMBs) are a key imaging marker of cerebral amyloid angiopathy (CAA), but their individual spatial distribution and potential clustering remain poorly understood.

Purpose

This study assessed whether CMBs exhibit patient-specific, flow territory-based clustering in sporadic (sCAA) and Dutch-type hereditary CAA (D-CAA). The arterial flow territory subdivision was chosen, considering the predilection of CMB location in the posterior cortex—not limited to the occipital lobe—and the prominent vascular role in CAA.

Study Type

Retrospective cross-sectional study.

Population

Thirty-three participants (17 (52%) female, 65 ± 12 years old, 19 sCAA; 14 D-CAA) without symptomatic intracerebral hemorrhage (sICH) and ≥ 1 CMB were included for the main analysis.

Field Strength/Sequence

3 T MRI susceptibility-weighted imaging, multi-echo 3D gradient echo.

Assessment

Spatial CMB patterns were evaluated by comparing observed versus expected CMB density within covariate-matched subgroups, accounting for age, sex, CAA-type, smoking, and hypertension. CMBs were manually annotated and assigned to six arterial flow territories (left/right ACA, MCA, PCA) using a standardized atlas. A sensitivity analysis was performed on all participants (including sICH+ participants).

Statistical Tests

Expected values were calculated by multiplying flow territory CMB proportions per covariate-matched subgroup. Unweighted individual chi-square statistics were calculated, and statistical significance was assessed via permutation testing (10,000 replications) per covariate group.

Results

Across covariate-matched subgroups, observed CMB counts significantly deviated from expected distributions, indicating substantial interindividual variability and within-participant clustering (χ² = 6920). The effect remained significant when including sICH+ participants (χ² = 102,027, n = 85; 44 sCAA; 41 D-CAA). CMB densities differed across flow territories, with highest median densities in the right PCA, right MCA, and left PCA (0.173 (IQR = 0.387), 0.121 (IQR = 0.238), and 0.113 (IQR = 0.389) counts per cm³, respectively).

Data Conclusion

These findings indicate CMBs in CAA are not identically distributed but show patient-specific clustering within flow territories, suggesting variable regional vulnerability, potentially influenced by local pathology or vascular anatomy.

Evidence Level

1.

Technical Efficacy

Stage 2.