For the reporting radiologist

ECG reporting that fits between the other lists.

Open a study and the measurement, the findings and the severity are there, with the evidence for each. Read it, disagree with it if you disagree, sign it — and move on to the next modality.

Role in the system: CardiologistHolds clinical authority

Request access How the pipeline works

What changes

What this means for radiologists, specifically.

Each item below corresponds to something in the build: a screen, a permission or a guard. Nothing here is a study result — this build has no clinical evidence base, and inventing one would be the worst possible way to sell it.

Measured before you look

The numbers are already there and already attributable to an algorithm version. Nothing needs to be counted off the paper before the clinical question can be considered.

Severity decides what you open first

The worklist puts the escalated traces at the top with their acknowledgement state, so the interruption is the one that mattered rather than the one that arrived.

Disagreement that is visible

A second read is a recorded act. Where it differs from the first, both conclusions stay in the study with their authors — which is the evidence a double-reporting policy actually needs.

Reports that leave cleanly

A signed study transmits to your PACS-adjacent systems over FHIR R4 or HL7 v2.5 with a delivery log for every attempt, so “was it sent?” is answerable from the record.

What it removes

The arrangement this replaces.

Stated as problems rather than statistics, because they are the ones a reader recognises from their own week.

ECGs are the interruption in a day built around images

Nobody reports ECGs continuously. They arrive in the middle of a reporting session, and the cost of switching to a trace that needs arithmetic before it can be read is paid by whatever was on screen before.

A signing queue is not a reading session

Much of what arrives is unremarkable and needs confirmation rather than deliberation. Mixed in with it are the traces that need a second look, and nothing in the list tells them apart.

Double reading is hard to demonstrate

When a study has been looked at by two people, the fact that the second disagreed is usually in a conversation rather than in the record.

Permissions held

What this role can do, read from the access-control table.

Generated from the same permission table the API enforces, so it describes access that exists rather than access that was described. If a permission is added to this role it appears here; if one is removed, it disappears.

Reporting ECGs is an interpreting-clinician job in this system, so a radiologist is given the same clinical role a cardiologist holds: run the analysis, review, override and sign.

Cardiologist

16 permissions held

See patient demographics: Name, hospital number and encounter details, for patients in this facility.
See the study list and its metadata: Which recordings exist, when they were taken, and what state each is in.
View the waveform: The stored samples themselves, rendered on calibrated ECG paper.
Read interpretations: Measurements, findings and severity for a study, including anything a clinician has written.
Run the analysis engine: Produce a new automated interpretation from the stored samples.
Record a clinical review: State agreement or disagreement with the automated interpretation, with a note.
Override a finding: Replace an automated finding with a clinician’s own, which is recorded as an override for ever.
Sign an interpretation: Attach a named, digest-bound signature to the exact interpretation it covers.
Re-run the analysis: Analyse a study again, for example after a newer algorithm version is approved.
Generate a narrative report: Produce the written summary, which describes validated findings and may not introduce new ones.
Transmit a signed study: Release a signed interpretation to another system over FHIR or HL7.
See critical alerts: The worklist of findings that need a person, and its acknowledgement state.
Acknowledge an alert: State that a person has seen it, so it is never left unattended.
Close an alert: Record how it was resolved, with a closure note.
Read operational analytics: Throughput, turnaround and worklist measures. Not a diagnostic accuracy figure.
Read the governance register: Which algorithm versions exist and under what intended use.

Screens this role opens

ECG Queue/queueReports/reportsWaveform viewer/ecgCritical Alerts/alerts

Navigation is filtered by permission rather than by role name, so a screen this role cannot open is rendered as a label explaining which permission it would need — rather than as a link that fails when it is clicked.

Separation of duties

What this role cannot do, and why that protects them.

Every access model is also a statement about accountability. These limits are the ones that make the audit trail worth having — and the ones that stop a signature meaning something it does not.

You cannot interpret a study the quality gate withheld; you can record a manual read, which is stored as a human conclusion rather than an automated one.
You cannot transmit an unsigned study. Transmission is the moment a clinical document leaves under a professional’s name, and the signature is what makes it one.
You cannot change the engine’s rules. Thresholds you own are your escalation policy; the measurement logic is versioned separately and governed by an administrative role.

Read together with the full pipeline

The limits above are enforced by the same rules that govern the pipeline described in how it works, and the regulatory position — including what this build does not establish — is stated on safety & limits.

Other audiences

This is one seat in the room.

The system is designed around the fact that the person who records an ECG, the person who reads it and the person accountable for the service are three different people.

Hospitals & clinics

What this is not

This is a demonstration-grade clinical system. It is not a medical device, it holds no regulatory clearance anywhere, and it has not been validated against a clinical dataset.

It does not diagnose and it does not replace clinical assessment. Every interpretation is decision support for a qualified professional, who reviews, overrides and signs it — and the record shows which professional.

No sensitivity, specificity, agreement or accuracy figure appears on this site, because none has been established for this build. Any such figure would be invented. Read the full position.

See it on your own recordings.

Send a request, and a person reviews it. If it is approved you get a link that creates your own organisation, with the first 2 analysed studies included and no card details — and you are its administrator from the moment you set your password.

Request access Who it is for