Oura Ring Gen 4 sensor data — not clinical measurementsN=1 case study — not validated for clinical decisionsHEV diagnosed Mar 18; interpret findings cautiously in this Day 15 post-ruxolitinib window

Dashboard

Oura Digital Twin

Generated 2026-04-06 · 2 patients · 17 analysis modules

Continuous wearable monitoring via Oura Gen 4. 17 analysis modules run daily across 2 patients in 2 countries, producing interactive HTML reports with clinical-grade signal processing, causal inference, comparative analysis, and predictive modeling. Drug response detection at p=0.01.

Core

Heart rate, HRV, sleep, activity, SpO2, and readiness trends across the full observation window.

Biomarker Trends

Composite biomarker indices combining multiple Oura signals into clinical-grade scores.

Sleep architecture, staging distribution, efficiency, and circadian rhythm analysis.

Clinical

Causal: Ruxolitinib

Bayesian causal impact and interrupted time-series analysis of ruxolitinib response.

GvHD Prediction

Hidden Markov and state-space models predicting GvHD flare probability from wearable signals.

SpO2 trend monitoring and bronchiolitis obliterans syndrome screening thresholds.

Advanced

Frequency-domain HRV, Poincare plots, DFA, sample entropy, and autonomic balance metrics.

Unscented Kalman Filter digital twin tracking latent inflammatory and autonomic states.

Foundation Model

Chronos foundation model forecasting with prediction intervals and anomaly scoring.

Anomaly Detection

Matrix Profile, Isolation Forest, and CUSUM anomaly detection across biometric channels.

Interactive 3D scatter of sleep, HRV, and activity with treatment phase coloring.

Comparative Analysis

2 PATIENTS · 2 COUNTRIES

Cross-patient analysis comparing Henrik (post-HSCT, Norway) and Mitchell (post-stroke, USA). Same wearable platform, different recovery trajectories. Drug response detection at p=0.01.

Ruxolitinib HRV Response

+17.4% HRV

p=0.005 · Statistically significant

Lowest HR Response

-5.6% bpm

p=0.006 · Drug-period effect

REM Sleep Gap

d=-1.26 Cohen's d

Large effect · Henrik deficit

Anomaly Similarity

0.52 cosine

Henrik: HR-driven · Mitchell: sleep-driven

Treatment Response Detection

Ruxolitinib response: HRV +17.4% (p=0.005), HR -5.6% (p=0.006), Avg HR -5.3% (p=0.006). Overall drug response p=0.01.

Autonomic Recovery

HRV and heart rate recovery trajectories. Comparative autonomic nervous system profiling across two recovery contexts.

Sleep Architecture

REM gap Cohen's d=-1.26, deep sleep d=-0.66. Sleep staging distributions and efficiency across patients.

Activity-Recovery Coupling

Both patients show negative coupling: more activity correlates with worse next-day readiness scores.

Anomaly Fingerprinting

Cosine similarity 0.52. Henrik's anomalies are HR-driven, Mitchell's are sleep-driven. Distinct recovery signatures.

Patient Overview

Side-by-side comparison of Henrik (post-HSCT) and Mitchell (post-stroke): HR, HRV, sleep metrics.

Context

Methodology, limitations, and honest assessment of what this system can and cannot do.

Planned analyses, validation targets, and next steps for the digital twin platform.

All metrics derived from Oura Ring Gen 4 consumer wearable data. Not clinical-grade measurements.

Two-patient case study (N=2). Not validated for clinical decision-making. Not a medical device.

Data: January 08 – April 05, 2026 · Post-intervention: 21 days (ruxolitinib, March 16) · Drug response p=0.01

Open source under MIT License · © 2026 The Educational Equality Institute · GitHub