Every modality has an explicit status: not done, done with interpretable findings, or attempted but uninterpretable.
UNVALIDATED RESEARCH PROTOTYPE — NOT FOR PATIENT CARE. Fusion training is synthetic/invented; rules, thresholds, and hospital registry entries are unreviewed test placeholders.
A transparent research workflow, not a clinical service
Native structured findings keep missing, uninterpretable, and observed information separate. Safety rules run before synthetic-trained model scores, and every suggested pathway remains subject to explicit research review.
Missing is not negative. Non-interpretable and non-done sections are cleared and cannot carry findings.
For eligible adults, deterministic safety rules establish a minimum level before learned classifiers combine findings and explicit missingness.
A model score is labelled as a model estimate only. The final level cannot be lower than a red-flag rule floor.
Age unknown causes abstention; paediatric age does not run adult diagnostic logic. No treatment, timing, or message sending is generated.
Destination routing uses fictional capability data and unverified test registry placeholders; drafts are displayed, never sent.
Six ranked conditions
- Stroke with large-vessel occlusion (LVO stroke)
- Intracerebral haemorrhage (ICH)
- Subarachnoid haemorrhage (SAH)
- Major trauma
- ST-elevation myocardial infarction (STEMI)
- Non-ST-elevation acute coronary syndrome (NSTE-ACS)