Patient Vitals Trend & Anomaly Detection

Cohort Stream

Vital sign

Patients in cohort 6

Anomaly Detector

Moving-average window 8 pts

Flag threshold |z| > 2.5

Live Readouts

Cohort mean

—

Cohort σ

—

Anomalies flagged

Trend (20 pts)

—

Patient trace (value over time) Rolling moving average Flagged anomaly (|z| over threshold)

How it works

Each patient's vital sign streams in as a noisy time series. For every new reading x_t, a trailing moving average and standard deviation are computed over the previous w points (a simple Shewhart-style control chart):

MA(t) = (1/w) · Σ x_(t-i),  i = 1..w
σ(t)  = sqrt( (1/w) · Σ (x_(t-i) − MA(t))² )
z(t)  = (x_t − MA(t)) / σ(t)

A reading is flagged as an anomaly when |z(t)| exceeds the chosen threshold — the same logic clinical monitoring systems use to catch a vital sign drifting outside a patient's own recent baseline, not just a fixed "normal range". A tighter window reacts faster but is noisier; a lower threshold catches more events at the cost of false positives.

Vital sign — switches the simulated baseline, noise profile and axis scale (heart rate, SpO₂, systolic BP).
Patients in cohort — how many independent traces stream at once, laid out along the depth axis.
Window / threshold — tune the detector exactly as in the formula above; anomaly markers update live.
Trend (20 pts) — the slope of a least-squares line fit through the cohort's last 20 mean readings, reported as rising, falling or stable.

Heart rate (bpm) t = 0

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.