Inflammation of brain blood vessels drives cortical superficial siderosis in cerebral amyloid angiopathy

A routine brain scan finding long dismissed as a passive scar of old bleeding may in fact be the visible tip of an active, treatable inflammatory process, according to new research that could reshape how specialists diagnose and manage one of the most dangerous small-vessel diseases of aging. The study, published in Annals of Clinical and Translational Neurology, used a specialized magnetic resonance imaging technique to reveal widespread inflammation of the brain’s lining and vessel walls in patients with cerebral amyloid angiopathy, a condition in which amyloid protein accumulates in arteries and leaves them brittle, prone to rupture, and capable of causing devastating brain hemorrhages.

Cerebral amyloid angiopathy, or CAA, has traditionally been viewed as a slowly progressive, largely non-inflammatory disease of older adults, driven by the deposition of beta-amyloid within the walls of cortical and leptomeningeal vessels. In a subset of patients, however, the immune system appears to mount an autoimmune response against the amyloid-laden vessel walls, producing a condition known as CAA-related inflammation. Current diagnostic criteria for this inflammatory syndrome require progressive neurological symptoms together with confluent, asymmetric white matter changes on MRI. But clinicians have increasingly encountered patients whose symptoms and imaging do not fit that narrow definition, prompting the proposal of a broader “CAA-related inflammation spectrum” concept. The new study provides some of the most systematic evidence yet that this spectrum is real, common, and detectable with the right tools.

Researchers at Otto-von-Guericke University Magdeburg in Germany examined a consecutive series of 15 patients with CAA-related cortical superficial siderosis, a condition marked by persistent deposits of hemosiderin, a blood breakdown product, along the surface of the brain. Siderosis is thought to arise from repeated, self-limiting bleeds from fragile, amyloid-ridden vessels just beneath the brain’s outer membranes, and it is one of the strongest imaging predictors of future fatal intracerebral hemorrhage. Between May 2022 and September 2025, the team identified 81 patients meeting probable CAA criteria by the Boston criteria version 2.0, of whom 15 had cortical superficial siderosis and underwent high-resolution vessel wall imaging as part of their diagnostic workup. Baseline characteristics such as age, sex, and hemorrhage prevalence did not differ between included and excluded patients, suggesting the sample was not strongly selected.

The imaging protocol was demanding. Each patient underwent a standardized 3-Tesla MRI on a Philips Achieva scanner a median of five days after clinical presentation, including T1-weighted, T2-weighted, FLAIR, T2*-weighted, and diffusion sequences, along with pre- and post-contrast “black blood” vessel wall imaging. The vessel wall imaging used a transverse, single-slab, non-selective three-dimensional turbo spin echo sequence with generalized autocalibrating partial parallel acquisition for acceleration and spectral attenuated inversion recovery for fat suppression, achieving whole-brain coverage at an isotropic voxel size of 0.8 millimeters. Gadolinium contrast was administered to highlight inflamed vessel walls, and all enhancement findings were cross-checked against pre-contrast sequences to minimize artifacts, with veins carefully distinguished from arteries based on anatomical trajectory, continuity, and morphology.

Two experienced senior neuroradiologists independently analyzed all scans, applying the Auriel criteria for CAA-related inflammation, the STRIVE-2 criteria for vascular lesions, and validated rating scales for white matter hyperintensities, atrophy, and microbleeds. Interrater reliability was substantial, with a Cohen’s kappa of 0.74 for the presence of leptomeningeal enhancement and good-to-moderate agreement for vessel counts. The researchers also constructed a composite MRI inflammation score, ranging from 0 to 4, by summing the presence of leptomeningeal and cortical vessel wall enhancement on vessel wall imaging with sulcal hyperintensities and inflammatory parenchymal edema on native FLAIR.

The results were striking. Although only four of the 15 patients, or 27 percent, met the current diagnostic criteria for CAA-related inflammation, 14 of them, or 93 percent, showed vessel wall enhancement or sulcal hyperintensities near their siderosis, frequently extending well beyond it. Fourteen patients displayed leptomeningeal vessel enhancement and 13 showed cortical vessel wall enhancement, while vasogenic edema was present in three. Among patients with vessel wall enhancement, sulcal hyperintensities were visible on native FLAIR in every case, without corresponding hemorrhage on T1-weighted images that would suggest subarachnoid bleeding as the explanation. Both vessel wall and sulcal enhancement were most pronounced in regions affected by siderosis, though ten patients exhibited diffuse enhancement throughout the brain, and enhancement of large-caliber arteries was seen in nine patients. Only one patient with disseminated siderosis showed no inflammatory process at all, underscoring the heterogeneity of the disease.

The inflammatory findings were not merely radiological curiosities. Lumbar punctures were performed in eleven patients, and all fell within the Alzheimer’s disease biomarker continuum when classified by the ATN framework of amyloid, tau, and neurodegeneration markers measured with automated immunoassays. More importantly, the albumin quotient in cerebrospinal fluid, a standard indicator of blood-brain barrier dysfunction, correlated strongly and positively with the MRI inflammation score, accounting for roughly 64 percent of its variance. Even when patients meeting formal CAA-related inflammation criteria were excluded, the correlation held. This convergence between fluid biomarkers and imaging suggests that the contrast enhancement reflects a genuine breakdown of the blood-brain barrier and an active meningovascular inflammatory process, rather than a scanning artifact or an incidental finding.

Perhaps the most clinically compelling observation came from follow-up. Seven patients with neuroimaging evidence of inflammation on vessel wall imaging received corticosteroid treatment, typically 0.5 or 1 gram of intravenous methylprednisolone per day for three to five days, and underwent follow-up MRI at a median of six months. All seven demonstrated a reduction in vessel wall enhancement and parenchymal edema, indicating that the inflammation is at least partially corticosteroid-responsive. One patient initially improved radiologically but later developed recurrent symptoms and renewed inflammatory changes, prompting a second course of steroids. Over a median clinical follow-up of about three months, one patient with an initial hemorrhage and vessel wall enhancement experienced a recurrent lobar hemorrhage before anti-inflammatory therapy could be started, a sobering reminder of the stakes involved in detecting and treating inflammation early.

The implications for diagnosis are significant. Eleven of the fifteen patients did not meet current criteria for CAA-related inflammation, which remain anchored solely on confluent white matter hyperintensities, yet the overwhelming majority of them nonetheless showed clear evidence of meningovascular inflammation. This echoes an earlier retrospective series of six patients with siderosis and transient focal neurological episodes, in which five of six showed inflammation on vessel wall imaging despite not fulfilling formal criteria. Taken together, these findings support the emerging concept of a CAA-related inflammation spectrum and argue for incorporating contrast-enhanced vessel wall imaging into the diagnostic toolkit. The authors suggest that fluid biomarkers reflecting endothelial activation, blood-brain barrier dysfunction, or specific inflammatory signaling pathways could complement imaging and further refine the boundaries of the spectrum, capturing atypical yet treatable presentations that the classical syndrome definition misses.

The pathophysiological picture emerging from this work is one in which recurrent microbleeds from amyloid-laden pial vessels deposit hemosiderin along the cortical surface, while a parallel inflammatory process involving the meninges and vessel walls contributes to further vascular fragility and bleeding susceptibility. If confirmed, this would transform cortical superficial siderosis from a passive marker of past bleeding into an active target for immunomodulatory therapy aimed at reducing vascular injury and hemorrhage risk, an unmet need in a disease for which treatment options remain limited. The authors caution that their study is retrospective, small, and lacks pathological confirmation of vascular inflammation, and that time-of-flight angiography was not routinely acquired to correlate findings with vascular anatomy. Prior work in CNS vasculitis, however, has shown that vessel wall imaging can reliably distinguish inflammatory from non-inflammatory vasculopathies, and early experience at 7 Tesla suggests that ultra-high-field imaging may further improve assessment of small-vessel inflammation. Contrast-enhanced FLAIR, which is more widely available, has also shown utility in detecting leptomeningeal enhancement in CAA, though it cannot directly visualize mural vessel enhancement. Larger prospective studies will be needed to validate these observations and determine whether anti-inflammatory treatment can genuinely improve long-term outcomes, including survival and hemorrhage risk, in patients with this under-recognized inflammatory face of cerebral amyloid angiopathy.