CancerXTwin

Simulation Lab · can the twin catch what usual care misses?

Data. Fully synthetic cohort generated in the browser, structured like Synthea FHIR R4 records (Patient, MedicationAdministration, Observation LOINC 8310-5 body temp, 751-8 neutrophils). No real patient appears here. Known issues are injected with ground truth, then the detection engine runs blind — so sensitivity and precision can be measured honestly. Real de-identified data (e.g. MIMIC-IV) needs credentialed access and a data-use agreement; Synthea bundles can be imported in the next build.

Cohort size

Injected issue rate (per patient, per issue type)

Random seed #{{ seed }}

Twin capabilities — switch off to see what each one is worth

{{ k.v }}

Detection performance by rule

Rule

Injected

Caught

Missed

False alarm

PPV

Sensitivity

{{ r.inj }}

{{ r.tp }}

{{ r.fn }}

{{ r.fp }}

{{ r.ppv }}

{{ r.sens }}

Patients with an issue or a flag · {{ listCount }}

84-day twin timeline · bars = reported temperature · teal ▲ = cycle given · amber ▲ = cycle held/delayed · violet = unconfirmed oral dose

Day 0Day 28Day 56Day 83

Day {{ e.d }}

Simulation for product validation only. Rules flag events for clinician review; they do not diagnose or change treatment. Thresholds: fever ≥ 38.0 °C within 14 days of infusion; neutrophils < 1.0 ×10⁹/L before a given cycle; cycle > 3 days later than scheduled without a documented reason; ≥ 3 consecutive unconfirmed oral doses.