Diagnostic Criteria and Rating Systems

1. First decide what job the instrument performs

2. STRIVE-2: describe what is visible

STRIVE stands for STandards for ReportIng Vascular changes on nEuroimaging. STRIVE-1 (2013) created a common vocabulary. STRIVE-2 (2023) preserved the phenotype-first principle, clarified evolution and mimics, and expanded attention to cortical microinfarcts, quantitative imaging, high-field MRI, vascular function, connectivity, and automated methods.

Core feature definitions in practical language

The most important STRIVE rule

Use descriptive terms without embedding an unproven cause. For example, report "strictly lobar cerebral microbleeds" rather than "CAA microbleeds" until the patient-level evidence is integrated. Report the sequence, field strength, slice thickness, echo time, lesion definition, rating instrument, and adjudication method because detection is sequence-dependent.

STRIVE-2 is not a replacement for STRIVE-1

The 2023 paper explicitly builds on the 2013 standard. A protocol should cite both when using original feature definitions plus newer guidance. It should predefine how new, disappearing, cavitating, or morphologically changing lesions are handled longitudinally.

3. Boston criteria: evolution and purpose

The Boston criteria began as clinical-pathologic criteria for CAA-related hemorrhage. Revisions progressively incorporated additional hemorrhagic markers and then non-hemorrhagic white-matter features:

• Original criteria: multiple lobar hemorrhages in an older patient, with other causes excluded.
• Modified Boston criteria (often called v1.5): incorporated cortical superficial siderosis as a hemorrhagic marker.
• Boston v2.0 (2022): lowered the age threshold to 50, broadened compatible presentations, and allowed severe centrum semiovale PVS or a multispot WMH pattern to combine with one lobar hemorrhagic lesion.

Boston v2.0 categories

The following is a research summary; use the published table and clinical expertise for formal application.

#### Definite CAA

Full postmortem examination shows:

• a compatible presentation (spontaneous ICH, TFNE, cSAH, or cognitive impairment/dementia);
• severe CAA with vasculopathy; and
• no other diagnostic lesion explaining the presentation.

#### Probable CAA with supporting pathology

Clinical data plus limited tissue from hematoma evacuation or cortical biopsy show:

• a compatible presentation;
• some CAA in the specimen; and
• no other diagnostic lesion.

Sampling is limited: a negative small specimen cannot reliably exclude segmental CAA.

#### Probable CAA by clinical and MRI data

All framework requirements apply:

• age 50 years or older;
• spontaneous ICH, TFNE, or cognitive impairment/dementia;
• no alternative cause of the hemorrhagic lesions; and
• no deep ICH or deep CMB on susceptibility-sensitive MRI.

In addition, one of these imaging combinations is required:

1. at least two strictly lobar hemorrhagic lesions in any combination of lobar ICH, lobar CMB, or cSS/cSAH foci; or
2. one strictly lobar hemorrhagic lesion plus one white-matter feature: severe centrum semiovale PVS or WMH multispot pattern.

Cerebellar hemorrhagic lesions are counted as neither lobar nor deep in v2.0.

#### Possible CAA

The same age, compatible-presentation, exclusion, and no-deep-hemorrhagic-lesion framework applies, plus either:

• one strictly lobar hemorrhagic lesion; or
• one qualifying white-matter feature.

Performance in the development study

For probable CAA, sensitivity and specificity varied by cohort and presentation. In the combined autopsy-standard group, reported sensitivity was about 74.5% and specificity about 95.0%. This is strong rule-in performance in the studied spectrum, not a universal constant. The criteria were derived from hospital-based, selected cases with compatible clinical presentations and tissue; performance can change in community screening, younger populations, underrepresented groups, low-quality MRI, or atypical presentations.

What Boston v2.0 does not establish

• It does not identify asymptomatic vascular amyloid before qualifying lesions appear.
• It does not prove that every counted lobar CMB was produced by CAA.
• It does not grade molecular burden or current activity.
• It does not incorporate amyloid PET, CSF, plasma, vascular reactivity, or diffusion metrics as formal criteria.
• It is not intended for hemorrhage with an evident alternative cause such as trauma, hemorrhagic tumor, vascular malformation, hemorrhagic transformation, or CNS vasculitis.
• The absence of deep hemorrhagic lesions is a classification requirement; mixed lobar and deep disease remains clinically and biologically important even when it falls outside probable CAA.

4. Edinburgh CT and genetic criteria

The Edinburgh model was developed for patients with lobar intracerebral hemorrhage, especially when MRI is unavailable or unsuitable. It uses:

• subarachnoid extension of the ICH on CT;
• finger-like projections from the hematoma; and
• APOE ε4 possession, when genotyping is available.

In the development cohort:

• no predictor present corresponded to a low predicted probability (about 7%);
• subarachnoid hemorrhage or APOE ε4 alone corresponded to an intermediate range (about 44-64%);
• subarachnoid hemorrhage plus either finger-like projections or APOE ε4 corresponded to a high probability (at least 95%).

The study used a small, population-based autopsy cohort and internal bootstrap validation. Its authors explicitly called for external validation. CT markers should be rated with a standardized instrument such as CHARTS, and acquisition/reconstruction matter.

Conceptual contrast with Boston

• Boston v2.0 addresses several CAA presentations and relies on MRI when no tissue is available.
• Edinburgh addresses the cause of a lobar ICH using acute CT morphology, optionally plus APOE.
• The instruments are complementary rather than competitors.

5. CAA-related inflammation criteria

CAA-ri is a potentially treatment-responsive inflammatory syndrome associated with vascular amyloid. Pathology spans predominantly perivascular inflammation and destructive transmural/intramural vasculitis; current guidance prefers CAA-ri as the umbrella term and reserves ABRA/CAA-related vasculitis for pathologically demonstrated angio-destructive disease.

Probable CAA-ri: the clinicoradiologic pattern

The validated 2016 framework requires:

• age 40 years or older;
• acute or subacute compatible symptoms such as headache, impaired consciousness or behavior, focal neurologic deficits, or seizures, not directly explained by an acute ICH;
• unifocal or multifocal corticosubcortical or deep WMH that is asymmetric and extends to immediately subcortical white matter, with asymmetry not explained by prior ICH;
• at least one corticosubcortical hemorrhagic lesion such as macrohemorrhage, microbleed, or cSS; and
• exclusion of neoplastic, infectious, and other causes.

In the original validation, probable CAA-ri achieved reported sensitivity of 82% and specificity of 97%; possible CAA-ri was less specific. Definite diagnosis requires tissue showing vascular amyloid plus appropriate perivascular, intramural, or transmural inflammation. Because disease is patchy, biopsy site selection matters and a negative biopsy may not exclude it.

Distinguish three related phenomena

Their overlap is mechanistically informative but they are not synonyms.

6. Neuropathology standards for arteriolosclerosis and CAA

B-ASC assessment

Common semiquantitative scales grade none, mild, moderate, or severe arteriolosclerosis, often in occipital white matter or multiple prespecified regions. Severe disease can include marked wall thickening, splitting/degeneration of the media, luminal narrowing or occlusion, and associated changes. Quantitative sclerotic indices measure wall-to-lumen geometry.

Minimum reporting should include:

• sampled regions and number of blocks;
• vessel compartment and size definition;
• stains and immunohistochemistry;
• grading rubric and whether the worst, mean, or regional score was used;
• interrater reliability and adjudication;
• whether CAA, inflammation, fibrinoid necrosis, mineralization, and atherosclerosis were separately rated;
• postmortem interval, fixation, section thickness, and image-analysis method.

CAA assessment

CAA should be characterized by compartment (leptomeningeal, cortical, capillary), region, severity, and vasculopathic change. Aβ immunohistochemistry and Congo red/polarization are complementary. Consensus protocols such as Love et al. and neuropathology data systems improve comparability, but regional sampling remains a fundamental limitation.

VCING

The Vascular Cognitive Impairment Neuropathology Guidelines integrate arteriolosclerosis, infarcts, hemorrhage, CAA, myelin loss, PVS dilation, hemosiderin leakage, and large-artery atheroma to estimate the likelihood that cerebrovascular disease contributed to cognitive impairment. VCING is an attribution framework, not a direct antemortem diagnostic test.

7. Visual rating instruments and burden scores

Microbleeds

• MARS (Microbleed Anatomical Rating Scale): records definite/possible microbleeds and their lobar, deep, and infratentorial locations; designed for reproducibility across observers and sequences.
• BOMBS (Brain Observer MicroBleed Scale): structured assessment developed around common disagreement sources and anatomic distribution.

Both improve measurement reliability. Neither converts location into pathologic certainty.

White-matter hyperintensities

• Fazekas scale: separate ordinal ratings for periventricular and deep WMH. It is fast and widely used, but coarse and subject to ceiling effects.
• Scheltens scale: more regionally detailed.
• Quantitative WMH volume is more continuous but sensitive to algorithm, training data, preprocessing, scanner, and manual correction.

Perivascular spaces

Visual scales commonly rate the number of PVS in the basal ganglia and centrum semiovale using the most affected slice/hemisphere and ordinal categories. Region matters: basal-ganglia PVS are often associated with non-amyloid arteriopathy, while severe centrum semiovale PVS support CAA in a compatible setting. These are group-level tendencies with overlap.

Total SVD score (general/non-amyloid-oriented)

The widely used 0-4 score assigns one point each for a qualifying burden of lacunes, CMBs, basal-ganglia PVS, and WMH. It is interpretable and useful for burden stratification but compresses heterogeneous lesions, thresholds continuous variables, weights each component equally, and changes meaning when components or cutoffs are modified.

CAA-SVD score

The 0-6 CAA-oriented score typically includes:

• 2-4 lobar CMBs: 1 point; at least 5: 2 points;
• focal cSS (up to 3 sulci): 1 point; disseminated cSS (4 or more sulci): 2 points;
• more than 20 centrum semiovale PVS: 1 point;
• confluent deep or irregular periventricular WMH meeting defined Fazekas thresholds: 1 point.

It summarizes visible CAA-associated burden; it is not the Boston diagnostic criteria. Avoid calling a modified score "the CAA-SVD score" without listing exact components.

8. A criteria application checklist for research

1. Define the eligible clinical presentation and age range.
2. Verify MRI contains an adequate susceptibility-sensitive sequence; record GRE versus SWI and parameters.
3. Exclude alternative causes before counting lesions toward CAA.
4. Map every hemorrhagic lesion as lobar, deep, infratentorial/cerebellar, or uncertain using a prespecified atlas/rule.
5. Distinguish cortical ribbon from juxtacortical and subcortical white matter when possible.
6. Rate cSS/cSAH foci independently from CMB count.
7. Apply severe CSO-PVS and WMH-multispot definitions exactly.
8. Record the final diagnostic category and each component so the classification is auditable.
9. Preserve an "unclassifiable/insufficient imaging" state; do not force missing data to negative.
10. If using pathology, model region sampled and MRI-to-tissue interval.
11. Report reader training, blinding, reliability, adjudication, and scanner/sequence mix.
12. Perform sensitivity analyses for uncertain lesions and alternate thresholds.

Key sources

• Charidimou A, et al. Boston criteria v2.0. *Lancet Neurology*. 2022. doi:10.1016/S1474-4422(22)00208-300208-3)
• Duering M, et al. STRIVE-2. *Lancet Neurology*. 2023. doi:10.1016/S1474-4422(23)00131-X00131-X)
• Wardlaw JM, et al. STRIVE-1. *Lancet Neurology*. 2013. doi:10.1016/S1474-4422(13)70047-870047-8)
• Rodrigues MA, et al. Edinburgh CT and genetic criteria. *Lancet Neurology*. 2018. doi:10.1016/S1474-4422(18)30006-130006-1)
• Auriel E, et al. CAA-ri criteria validation. *JAMA Neurology*. 2016. doi:10.1001/jamaneurol.2015.4078
• Gregoire SM, et al. MARS. *Neurology*. 2009. doi:10.1212/WNL.0b013e3181c34a7d
• Cordonnier C, et al. BOMBS. *Stroke*. 2009. doi:10.1161/STROKEAHA.108.526996
• Skrobot OA, et al. VCING. *Brain*. 2016. doi:10.1093/brain/aww214