Patent Portfolio

Cardiac Diagnostics

A device-invariant, interpretable ECG review layer assesses a patient's cardiac dynamical state from as little as a single lead. The method is training-free, built from first-principles signal geometry rather than a model fit to a labeled dataset, and is designed to abstain and flag for human review when its own confidence is below threshold, rather than force a guess.

• Device-invariant assessment across single-lead wearable and full clinical 12-lead recordings, with no retraining step
• Calibrated abstention: the system declines to render an opinion on low-confidence cases and logs them for human review
• Training-free, interpretable signal geometry rather than a black-box model
• Filed provisional, research-stage; not FDA cleared or submitted, not deployed

Molecular Phenotyping

An objective, blood-based molecular stratification signal targets a condition still diagnosed today by symptom criteria and exclusion. The method is built on real cohort data, and the work is honest about which sub-signals hold up under scrutiny and which do not.

• Objective, blood-based stratification signal, not a symptom-criteria checklist
• Built on real cohort RNA-seq data, with sub-signal-level honesty about what replicates
• Filed provisional, research-stage; not deployed or clinically validated

Sepsis Monitoring

A white-box host-response signature from a single blood draw, paired with an interpretable trajectory-review layer, tracks whether a patient is improving or worsening. The tool is offered alongside clinical judgment, never as a replacement for it.

• Single blood draw host-response signature
• Interpretable trajectory review that tracks improving vs. worsening course
• Filed provisional, research-stage; decision-support only, not an autonomous diagnostic

Early-Warning Detection of Critical Transitions

An interpretable, physically-grounded early-warning monitor tells you honestly when it can and cannot see trouble coming. It corroborates classical critical-slowing-down theory and is positioned as an auditable governance and fallback layer for high-stakes decisions currently made by an opaque model alone, not as a claimed improvement over deep-learning baselines.

• White-box monitor: every indicator traces to a physically meaningful signal property
• Honest, null-tested characterization of when enhancements do and do not add discriminative content
• Filed provisional, research-stage; not deployed, not a demonstrated superiority claim over deep learning