Technology

The Biological Signal Disorder Framework.

Regenesis-OS reads the structure of physiological signals, not just their values. Collapse is preceded by a measurable loss of order across the vital signs hospitals already monitor.


What We Measure

Three signatures of instability.

We quantify how organized a patient's physiology is, moment to moment. As control degrades, three properties move together, well before any single number crosses an alarm threshold.

Entropy

Rising disorder

Healthy regulation produces structured, predictable signals. As systems destabilize, that structure dissolves into randomness. We track the climb in signal entropy as an early marker of failing control.

Variance

Widening spread

Beat to beat and breath to breath, dispersion increases as the body loses its ability to hold a stable operating point. Growing variance signals reserve being exhausted.

Coordination

Loss of coupling

In health, heart rate, respiration, and blood pressure move in concert. Disorder shows up as a breakdown in that coordination across vital signs, ahead of overt deterioration.

The specific mathematics that turn these properties into a calibrated risk estimate are proprietary. What matters clinically is the result: a probability that strengthens as an event approaches.


The Early-Warning Curve

Accuracy strengthens as the event approaches.

The disorder signal sharpens in the hours before collapse. Performance is strong far out and becomes decisive close in, giving clinicians a widening, then tightening, window to act.

  • 0.91at 2 hours before
  • 0.82at 10 hours before
  • 0.72at 24 hours before

AUC, the area under the ROC curve, summarizes how well the model separates patients who will deteriorate from those who will not. 1.0 is perfect, 0.5 is chance.

Model AUC vs. hours before event

Hours before event


Deployment

Built on data you already collect.

The model runs on routine vital-sign streams from existing monitors. There is no new bedside hardware to buy, install, or maintain. It reads the signals already flowing through the unit.

No new bedside hardware. Software layered onto the data your monitors already produce.

Calibrated probabilities. Output is a clinically interpretable risk, suitable for direct display.

Bias-audited pipeline. Audited end to end for five distinct sources of bias.


Validation Cohort

Held out, matched, and tested honestly.

Performance was measured on a cohort of 1,258 ICU stays, matched on care unit and timing, and held out throughout development. External results came from separate hospital systems in the eICU database.

1,258
ICU stays in the validation cohort
0.77 / 0.82
External / internal AUC, in agreement
100%
Held out throughout development

See where this is going.

We are opening hospital pilot partnerships and an investor conversation.