Physiological Disorder Detection
AI-powered physiological disorder detection for ICU patients, identifying cardiac arrest approximately 10 hours before collapse.
The Problem
Roughly 290,000 people suffer in-hospital cardiac arrest every year. Existing early-warning systems detect deterioration too late.
Rule-based threshold alarms generate false positives and miss the real signal. The problem is not monitoring. It is what we measure.
The Solution
Biological systems show measurable disorder, rising entropy, increasing variance, loss of signal organization, hours before catastrophic collapse.
Regenesis-OS detects that disorder window. Not thresholds. The instability itself.
Signal randomness climbs as physiological control degrades.
Vital-sign dispersion widens before any threshold is crossed.
Coordination across vital signs breaks down ahead of collapse.
0.77 AUC
Externally validated across separate hospital systems (eICU).
0.82 AUC
Internal performance, with internal and external agreement.
10 Hr
A statistically significant improvement over the SOFA standard-of-care score at 10 hours pre-event.
Calibrated
Probabilities suitable for clinical display, not raw scores.
5 Sources
A pipeline audited end to end for five distinct sources of bias.
Read the full technical framework.
Explore Technology ›Give clinicians the hours they need to act before collapse.