Each vital sign is generated by an Ornstein–Uhlenbeck process relaxing toward a set point μ with physiological noise:
dX = θ(μ − X)·dt + σ·√dt·N(0,1)
"Inject Deterioration Event" ramps μ for HR/RR/temperature up and SpO₂ down over ~40 simulated seconds via a smoothstep curve — a sepsis-like decompensation reaching the wearable continuously, not just at ward-round check intervals.
The AI early-warning score aggregates NEWS2-style banded sub-scores (0–3 per vital) into a base score, then adds a trend term — the smoothed rate of change of that base score — the actual advantage of continuous AI monitoring over a periodic manual score:
R_base = Σ subscore(HR) + subscore(SpO₂) + subscore(RR) + subscore(Temp) [0..12]
R_AI = clamp(100·R_base/12 + 10·κ·slope, 0, 100) [slope = pts of R_base per sim-second]
t_crit = (100 − R_AI) / max(slope_per_min, ε) [minutes to projected critical]
Note: the reference 3D twin's own write-up states R_AI = 100·(R_base/12 + κ·dR_base/dt), which implies a ×100·κ trend coefficient — but its actual code applies trendTerm·10 with trendTerm = κ·slope, i.e. a ×10·κ coefficient (10× smaller). This 2D engine reproduces the code's real, numerically-checked behaviour (×10·κ) rather than the overstated prose formula, and exposes κ as a live slider so you can feel the trend term's actual weight directly.
- Baseline sliders — set the patient's resting physiological set point for all four vitals.
- Trend sensitivity κ — how strongly the rate-of-change of the vitals (not just their current value) pulls the AI score up; raise it to see the score cross alert thresholds earlier during a deterioration.
- AI trend term ON/OFF — toggle to compare the AI-augmented score against a plain static NEWS2-style score with no trend awareness.
- Inject Deterioration Event — starts a 40s (sim-time) decompensation ramp; watch the AI core in the network diagram shift from green to red before any single vital alone would trigger a hard alarm.
- Network diagram — drag to pan, scroll/pinch to zoom; sensor nodes stream data packets into the central AI core, whose color, size and pulse rate all track the live risk score.
- Speed — compresses/stretches simulated time relative to real time.