Worked Reasoning Cases

These cases teach evidence-calibrated reasoning. They are educational examples, not patient-specific clinical advice.

UC001 | Lobar ICH with disseminated cSS

Scenario: An older adult presents with spontaneous lobar ICH, multiple strictly lobar microbleeds, and disseminated cSS.

Reasoning: The phenotype strongly increases CAA probability in the intended symptomatic context. Apply current criteria, confirm sequence adequacy, and separately estimate recurrence and treatment risk.

Do not conclude: Do not infer that every cognitive symptom is caused by CAA or that competing pathology is absent.

UC002 | Cognitive presentation without hemorrhage

Scenario: A memory-clinic patient has WMH, CSO-PVS, and no hemorrhagic marker.

Reasoning: Boston v2.0 transportability is weaker in this spectrum. Consider AD and mixed pathology, verify MRI quality, and treat any CAA inference as uncertain.

Do not conclude: Do not use criteria performance from lobar ICH cohorts as if it were screening performance.

UC003 | Mixed lobar and deep microbleeds

Scenario: MRI shows both lobar and deep microbleeds with extensive WMH.

Reasoning: Model mixed or competing pathologies. Fine-grained lesion location and cSS may help, but neither distribution alone supplies a pathology diagnosis.

Do not conclude: Do not force a binary CAA-versus-hypertensive label.

UC004 | High ARTS score

Scenario: A research participant has a high ARTS score and worsening cognition.

Reasoning: Separate model output, pathology prediction, technical repeatability, outcome association, and clinical utility. Inspect cohort similarity and calibration.

Do not conclude: Do not label the participant pathology-positive without validated context and reference evidence.

UC005 | Subacute encephalopathy with asymmetric FLAIR

Scenario: An older adult develops seizures and cognitive change with asymmetric parieto-occipital FLAIR hyperintensity and lobar microbleeds.

Reasoning: CAA-ri is important, but infection, neoplasm, PRES, vasculitis, and seizure-related change require assessment. Treatment exposure determines whether ARIA enters the differential.

Do not conclude: Do not equate an ARIA-E-like appearance with one etiology.

UC006 | Positive amyloid PET in suspected CAA

Scenario: An older patient with cognitive impairment has positive amyloid PET and a small lobar microbleed burden.

Reasoning: PET confirms fibrillar amyloid but not its vascular compartment. Integrate regional PET, tau/CSF, MRI pattern, and mixed-pathology probability.

Do not conclude: Do not call a positive scan diagnostic of CAA.

UC007 | Reduced CVR with WMH progression

Scenario: A research participant has low CVR and increasing WMH over one year.

Reasoning: This supports vascular dysfunction as a risk marker for tissue progression, while etiology and causal reversibility remain unresolved.

Do not conclude: Do not infer CAA or arteriolosclerosis specificity from CVR alone.

UC008 | Radiology-report candidate phenotype

Scenario: A report mentions WMH, chronic infarcts, and deep microbleeds without direct image review.

Reasoning: Use the report for candidate generation and triage. Confirm sequence availability, negation, image-level location, and pathology reference before model development.

Do not conclude: Do not turn report text into validated diagnostic criteria.