How it works

How it works

The user uploads a study (the recorded electrocardiogram and the clinical context) — the pipeline builds a draft conclusion — the physician agrees with it and authorizes it with a qualified electronic signature — the document is formed as a structured medical document and goes for registration. An ECG recording made at home works in the same way: it reaches the physician already read and shaped into a conclusion, not raw.

Request a pilot Limits and responsibility

The path

The path of a study: from the ECG signal to the registration of the documentEight links joined by arrows in two rows. First row: a signal of any length — from thirty seconds to a 24-hour Holter recording; the measurement of waves, intervals, axis and rhythm; a term from the dictionary of the cardiology dialect; a versioned rule. Second row: a finding with its grounds; a structured medical document; the physician's signature made with a qualified electronic signature — the only link with a highlighted frame; registration in the registry. Below the links a reminder: every link carries a rule version, a signal window and a recording-quality assessment.THE PATH OF A STUDYSignalfrom 30 secondsto a 24-hour HolterMeasurementwaves, intervals,axis, rhythmTermdialect dictionaryRulewith a versionFindingwith its groundsDocumentstructuredSignaturequalified electronic,of the physicianRegistrationin the registryEvery link carries a rule version, a signal window and a recording-quality assessment

The chain is produced at a click. The evidence chain answers the question “what is the output based on”: behind a finding stand a versioned rule with a reference to a quotation from a trusted source, and the signal window from which the output was drawn — not the weights of a model.

Architecture

Architecture

The chain that assembles a decision: a single knowledge graph — the knowledge compiler — validation of the rule package — execution by a deterministic C++ computation core — the synthesis of reports — the physician’s workplace. The rules are defined in the graph in a rule-description language and compiled into executable code: a short response time and low operating costs, with no infrastructure for serving and running models.

The output is a structured medical document signed with the physician’s qualified signature under the requirements in force, with the signature verified against Russian national cryptographic standards and with the registration states tracked in the registry of medical documents.

The core

The decision core

Interpretation is run not by an AI model but by a deterministic pipeline following verified rules bound to quotations from trusted sources.

The decision core: four analysis tasks and the line beyond which the rules and the physician decideAt the top, four analysis tasks with an estimate of their difficulty: recording quality and artifacts — low; morphology, that is waves, intervals, the axis of the heart and rhythm — medium; ruling out plausible mimics of ischemia — high; the draft text of the conclusion — medium. Below a dashed line, the decision core in two parts: deterministic rules from the knowledge graph, each with a version and an author, and the authorization by the physician, who confirms or corrects the draft. At the bottom two lines: there are no trainable models in the decision path; the data do not leave the infrastructure of the organization or a cloud on Russian territory.ANALYSIS TASKSRecording quality and artifactslowMorphology: waves, intervals, axis, rhythmmediumRuling out plausible mimics of ischemiahighThe draft text of the conclusionmediumbeyond this: rules and physician decideTHE DECISION CORERules from the knowledge graphversion · author · producibleAuthorization by the physicianconfirms or correctsThere are no trainable models in the decision path.The data do not leave the loop of the organization or a cloud in Russia.

The hardest task is solved not by an algorithm but by architecture: not to “guess right”, but to recognize the features reliably, to state the grounds explicitly and to present them to the physician or the patient as an evidence chain.

The finding

What stands behind a finding

What opens up behind one finding of the conclusionAt the top, one finding in the conclusion, set off by a frame. An arrow downwards labelled “opens at a click” leads to four cards: the versioned rule; the signal window, that is the stretch of the recording from which the output was drawn; the dictionary term the finding is bound to; the recording-quality assessment. At the bottom a line: what can be produced is the rule, not the weights of a model.A finding in the conclusionone line of the protocolopens at a clickRulewith a versionSignal windowpart of the recordingTermwhat it is bound toQualityof the recordingIn a dispute you can produce the rule with its evidence chain, not the weights of a model

Every finding opens into grounds formed from the time window of the signal and the features computed over it. What a rule computes is a numeric value, a finding or a wording in the draft conclusion. The physician or the patient receives an assembled document in which every wording can be checked.

Two loops

Two loops of one product

Two loops: an ECG recorded at home and the physician's workplace meet at the physician's authorizationTwo columns. On the left, on the patient's side: an ECG recorded at home with the person's own device. On the right, at the functional-diagnostics physician: a study taken in the room. Both electrocardiograms enter a common pipeline: measurement, term, rule, finding. Further down, on the left the patient receives a case summary and a conclusion as an informational draft; on the right the physician receives a draft conclusion ready by the time they sit down to the case. At the bottom a common band: the physician authorizes the document, and the document goes into the patient's record and on to registration.ON THE PATIENT'S SIDEAT THE FUNCTIONAL-DIAGNOSTICS PHYSICIANA recording at homeown device, on the person's own initiativeA study taken in the roomthe flow of screening and appointmentsThe common pipelinemeasurement · term · rule · findingCase summary and conclusionan informational draftDraft conclusionready by the time the case is reviewedThe physician authorizes — the document goes to the record and the registry

An ECG recording made at home reaches the physician not as a photograph of the strip but as an analyzed case. In return the patient receives the document in their own record. Neither of the two surfaces gives this on its own — what works is their integration.

Dictionary

The dictionary of the cardiology dialect

The place of a training set is taken by a curated dictionary. As of the measurement date the dictionary holds 111 descriptors across seven layers of analysis: signal, quality, waves, morphology, rhythm, clinical conclusions, patterns; bound to the electrocardiographic subsets of international terminologies. Quality is checked term by term rather than declared over a sample.

The second mechanism is the accumulation of knowledge through the authorization of conclusions: the physician’s comment becomes a de-identified remark and lets a specialist correct the rule on stated grounds, with an author and a version. The third mechanism is the perimeter: the data do not leave the infrastructure of the organization or a cloud on Russian territory, and they do not go into the training of third-party models.

The way in is a pilot with a before-and-after measurement.

Request a pilot For clinicians
HealthOS Cardio is not a medical device. The machine does not make a diagnosis and does not sign the document: a physician authorizes the draft conclusion with a qualified electronic signature.