Rospatent: decision to grant a patent on machine-readable representation of medical knowledge

Patent granted

On 26 May 2026, the Federal Service for Intellectual Property (Rospatent) decided to grant a patent for the invention “A method of machine-readable representation of medical knowledge” under application No. 2025138295/28 (086558) filed on 25 December 2025. The invention has been granted patent No. 2865309.

Details

Field Value
Office Federal Service for Intellectual Property (Rospatent)
Decision date 26 May 2026
Invention title A method of machine-readable representation of medical knowledge
Application number 2025138295/28 (086558)
Patent number 2865309
Filing date 25 December 2025
Status Decision to grant the patent

A decision to grant is a stage of state examination; final patent identifiers are subject to Rospatent registration and publication records.

Why it matters for HealthOS

The invention defines a formal method for turning medical knowledge into a machine-readable form suitable for validation, compilation, release, and bedside execution. For the HealthOS platform — including HealthOS ICU — this is not a standalone “patent slide”, but the foundation of the scientific and technical contribution: from authorized semantics to a deterministic execution process with lineage receipts and a closed refinement loop.

In short, the platform is built on compiling clinical knowledge into verifiable, deterministic runtime code, not as an AI black box used at the bedside.

Five contribution areas

Medical informatics. Executable knowledge travels from an authorized knowledge-graph snapshot through compilation and release to runtime evidence and graph refinement — a closed loop without parallel “sources of truth”. Instant health assessment in the operating room and ICU is reproducible derivation of diagnostics and bedside commands from data streams, instead of opaque AI inference.

Executable knowledge and OR/ICU → · Closed-loop governance →

Semantic gateway. A unified semantic projection of device telemetry, clinical records, and coded terms onto the patient and care-context model, through a vendor-neutral edge layer (HL7 v2/MLLP, IHE PCD, ISO/IEEE 11073 SDC; FHIR R4 as a projection of the audit log). Data integration, rule execution, and governed write are separate planes.

Standards and interoperability → · Reference architecture →

Knowledge compiler. A medical knowledge graph with SHACL validation and role-based governance (author → approver → compiler → release operator) produces a compiled bundle — a deterministic operational projection of an approved snapshot — with end-to-end term traceability from ontology to the executable module.

Compiler languages → · Term traceability → · Authorization and diagnostic contour →

Software systems engineering. Compute core and safety kernel are separate deterministic contours; stream and batch analyzers work on a mix of oscillograms and named features through a deterministic signal-evidence pipeline; governed release from a trusted snapshot holds a continuous chain of trust to executable modules.

Compute core and safety kernel → · Governed release → · Chain of trust →

Clinical application novelties. Deterministic supervisory control for OR/ICU traces recommendations and commands back to an approved rule through a vendor-neutral supervisory layer; medical-device control rolls out in phases, with three command classes (green / yellow / red) and mandatory confirmation for high-risk actions. The knowledge graph evolves through a governance cycle, not AI model drift.

Full OR/ICU concept → · Safety kernel → · Phased rollout →

The patented method underpins not only HealthOS ICU but adjacent HealthOS contours where clinical semantics compile into verifiable informational and supervisory artifacts — while preserving each product line’s scope and regulatory positioning.

This announcement is not a claim of clinical efficacy, completed medical-device registration, or commercial availability of every capability described. Verifiable statements about implemented, partial, and out-of-scope capabilities are in the conformance matrix.

Documents