CTRL · STANDBY
Per-Patient Twin · Diabetes Module

The patient is the reference.

Precision medicine in its purest form.

A hash-locked, closed-form digital twin. Glycaemic state, velocity, and time-in-range tracked against the patient's own trajectory — not against a cohort average. Hypoglycaemia signal rising 90 minutes before clinical onset.

CTRL
σ-ENGINE
Σ-STATE
0.0
ΔG/Δt
0.00
TIR
64
Σ-ALERT
0
Exemplar trace · Pre-hypoglycaemic window
HUPA-UCM · 5-min CGM · N=24 · AUROC 0·9938 · 90-min median lead
TARGET 70–180 mg/dL HYPO 70 GLUCOSE σ-STATE σ-ALARM FIRES HYPO EVENT 90-MIN MEDIAN LEAD −3h−2h−1h0
The 24-hour view

Continuous monitoring. Daily glycaemic management.

Same patient. Same engine. Full diurnal cycle with meal boluses, postprandial excursions, and nocturnal hypoglycaemia anticipation. The σ-state runs continuously — visible alerts only when warranted.

Daily trace · CGM + insulin + σ-warning
T1DM · 5-min CGM · 4 boluses · 1 σ-alert · 1 hypo prevented
TARGET 70–180 mg/dL σ-ALERT · 04:30 90-MIN LEAD PROJ HYPO · 06:00 HYPO 70 GLUCOSE (mg/dL) 15g CHO 8u 10u 6u 12u 00 04 08 12 16 20 24 280 240 180 140 90 70 50

At 04:30 the σ-state crossed threshold — 90 minutes before glucose would have crossed 70 mg/dL. Predictive 15 g rescue carbohydrate at 05:30 averted the event; glucose nadir at 06:00 stayed above the hypoglycaemia threshold. The same engine that runs the 4-hour pre-event exemplar above runs continuously across the 24-hour cycle without retuning.

The Range

Five domains. One SHA-locked engine.

The same controller, same parameter file, no per-domain retuning. The diabetes twin above is one instance of five. Each curve below shows the σ-trajectory or characteristic pattern for the corresponding domain, with the peak performance metric on publicly accessible data.

ICU
Haemodynamic
deterioration
EVENT 6–18 H LEAD
0·900
AUROC
N=2,492 ITT · eICU-CRD
vs APACHE-IV 0·727
AF
Paroxysmal
pre-onset prediction
AF ONSET 21-MIN MEDIAN LEAD
0·925
AUROC · 21-min lead
PhysioNet AFPDB
Paroxysmal onset prediction
PD
Parkinson's
gait architecture
CONTROL PD GAIT STRIDE AMPLITUDE
0·895
AUROC LOOCV
0·884 LOCOCV
3 independent cohorts
OSA
Sleep apnoea
ECG-derived screening
σ RISES AT EACH APNEA
0·984
Patient-level AUROC
PhysioNet Apnea-ECG
Per-patient aggregation
The Extension

From patient to device.

The same N=1 architecture applied to a medical device's own operational signals rather than to a patient's physiological signals yields per-device drift detection — denoted σ_machine.

Slow longitudinal degradation below conventional alarm thresholds becomes visible at the per-device scale. The same engine that governs the patient governs the device delivering the therapy.

The Architecture

Three layers per patient.

The same closed-form engine, hash-locked, applied identically across every domain. Hill-equation kernel calibrated from the patient's own pre-perturbation trajectory.

Layer 01
Discriminatory state
Per-channel standardised deviations against the patient's own rolling baseline pass through an individual Hill saturation function. A multi-channel concordance gate produces the state classification. The patient is the reference; no population norm.
Layer 02
Oracle drift
Geometric deviation of the observed trajectory from the patient's own maximum-possible trajectory — the Oracle Line — combined with Oracle-shift. Detects divergence from the individual ideal, rather than from a cohort mean.
Layer 03
Counterfactual governance
For each candidate intervention, predicted Oracle drift is computed and the intervention minimising future drift is selected. An actionable recommendation, not just a probability.
Origins

Clinical pharmacology, applied at the individual scale.

The architecture originates in the nifedipine rate-of-rise pharmacokinetic studies at the Leiden school in the mid-1980s — specifically, the observation that the velocity of physiological change governs tolerability independently of the magnitude of the dose itself.

Generalised, that observation says: the rate at which a system moves matters as much as where it is. Per-patient analytics means giving each patient their own reference trajectory and watching the rate at which they depart from it — rather than scoring them against a population average that does not exist in their physiology.

Four decades of clinical pharmacology and direct regulatory experience across FDA, EMA, MHRA and PMDA stand behind the architecture.

Technical notes

The science, in detail.

Two short technical notes set out the conceptual foundation and the statistical framework underlying the Zyvolance architecture. Written for clinicians, regulatory readers, and methodologists. Institutional voice; no marketing.

Engagement

By introduction only.

Zyvolance engages with regulated medical-device manufacturers, clinical-research institutions, and regulatory bodies on a confidential basis. Serious enquiries are welcomed at the addresses below.

Correspondence