The product
This is the software, not a description of it.
Not a medical device. Not approved for diagnostic use.
Every figure and waveform on this page is generated by this system’s own synthetic ECG generator from the profiles the demonstration deployment serves. No patient data is involved, and none of it is a clinical finding. The software has not been validated against a clinical dataset, holds no regulatory clearance, and its output is decision support for a qualified professional rather than a diagnosis. The full position, including what this build does not do.
The work queue
Opened by: cardiologist, radiologist, physician, technician
The screen a reading session starts in. Studies arrive already measured and already graded, and the order is severity first, then quality, then age — so the trace that matters is not behind the one that merely arrived first. A recording whose quality was withheld appears with nothing assessed for it, rather than an optimistic "normal".
ECG work queue
1 469 studies · page 1 of 37- Critical findingFixture Patient 04MRN-…4417 · Emergency Department2026-02-01 09:05Ventricular tachycardia · 178 bpmCritical — immediate attention
- Critical findingFixture Patient 11MRN-…1180 · Coronary Care Unit2026-02-01 08:58ST-segment elevation, anterior leads · 121 bpmCritical — immediate attention
- UrgentFixture Patient 07MRN-…9032 · Cardiology Outpatients2026-02-01 08:12Prolonged QTc interval · 44 bpmUrgent review recommended
- Not assessedFixture Patient 19MRN-…2264 · Pre-operative Assessment2026-02-01 09:06Interpretation withheld — quality below the validated thresholdInterpretation withheld
- RoutineFixture Patient 02MRN-…6673 · General Medicine2026-02-01 08:40No significant abnormality detected · 68 bpmNo significant abnormality detected
Sorted by urgency, then by signal quality, then oldest acquisition first. Studies awaiting analysis are shown first, because nothing has been assessed for them.
The interpretation
Opened by: any role that can read a report
Every measurement carries its reference range, a confidence figure and the leads it was derived from. Where the signal does not support a measurement, the field says not measurable rather than estimating one — which is why the PR interval in this example is empty. Nothing here is produced by a language model, and the reporting model is recorded on the screen as none.
Automated interpretation
VOMEDIQ Deterministic ECG Analysis vomediq-analysis-1.6.1 · rules 1.0.0Potentially critical finding — immediate clinical attention required
Overall confidence 0.58 · 5 findings reported; the most significant is listed first.
Rhythm
Ventricular tachycardia
Rhythm confidence
0.62
Processing time
142 ms
Reporting model
none
Measurements
| Measure | Value | Reference | Confidence | Derived from |
|---|---|---|---|---|
| Heart rate | 178 bpm | 60–100 | 1.00 | II |
| PR interval | not measurable | 120–200 | — | II |
| QRS duration | 144 ms | 80–120 | 0.94 | II, V1 |
| QT interval | 312 ms | 350–440 | 0.81 | II |
| QTc (Bazett) | 412 ms | 350–450 | 0.81 | II |
| QRS axis | leftward | −30° to +90° | 0.88 | I, II, III |
| R-wave progression | poor | progression by V4 | 0.86 | V1–V6 |
A measurement the recording does not support is reported as not measurable rather than estimated. Nothing on this screen is produced by a language model: the reporting model is recorded as none, and the narrative layer, where a deployment enables it, is given validated findings as its only input.
The findings
Opened by: cardiologist, radiologist, physician
Findings are ranked by how much they matter, and each carries the measurement that supports it and the rule version that produced it. The paragraph under "what this cannot establish" is part of the finding rather than a disclaimer bolted to the page: a system that states a ventricular origin it cannot actually prove has told the reader something false.
Findings
Each carries the measurement behind it and the rule that produced it- Critical — immediate attention
Broad-complex tachycardia consistent with a ventricular origin
VENTRICULAR_TACHYCARDIA_PATTERNleads II, V1, V6measured QRS 144 ms at 178 bpm- Measured evidence:
- Ventricular rate measured 178 bpm with a QRS duration of 144 ms. No organised atrial activity was identified preceding each complex.
- What this cannot establish:
- The recording does not reliably distinguish ventricular tachycardia from a supraventricular rhythm conducted with aberration. Haemodynamic state is not represented anywhere in this recording and is central to the decision.
- Abnormal — review required
Widened QRS with a left bundle branch pattern
LEFT_BUNDLE_BRANCH_PATTERNleads V1, V6, Imeasured 144 ms- Measured evidence:
- QRS duration measured 144 ms, at or above the 120 ms limit for a normal QRS. Net QRS deflection: V1 −0.29 mV, V6 0.26 mV.
- What this cannot establish:
- A new left bundle branch pattern may accompany acute myocardial infarction, and the usual ST-elevation criteria do not apply. Only a previous ECG separates them.
“What this cannot establish” is part of the finding, not a disclaimer added to the page. A finding with no stated evidence is a defect, not a display option.
The waveform
Opened by: cardiologist, radiologist, physician, nurse, technician
Plotted from the stored samples with every point in order and straight lines between them. No smoothing and no decimation, because both change what a sharp QRS looks like and the high-frequency content is exactly what an interpretation turns on. The grid is drawn from the same scale, so a count of boxes on screen is a count of boxes on paper — and the 1 mV calibration pulse at the head of each row makes the gain checkable by eye.
ECG waveform — recorded samples, unmodified
vomediaq viewer · every sample, in order, joined by straight linesNo smoothing
Notching, RSR′ patterns and pacemaker spikes are the high-frequency content an interpretation turns on.
No decimation
Plotting every sample is why a sharp QRS stays sharp when the view is zoomed.
Verifiable gain
A 1 mV calibration pulse heads every row, so the scale is checkable by eye.
The alert worklist
Opened by: cardiologist, physician, nurse
A finding that meets your escalation threshold becomes an alert with an owner, an acknowledgement state and a target time. The overdue row is the reason the screen exists: a critical finding with nobody’s name against it shows up as exactly that, instead of living in somebody’s memory of a phone call.
Critical alerts
Raised by the deterministic rule engine, evaluated against the stored interpretation- Critical — immediate attentionSeverity 3 at publication2 h 14 mAcknowledgedDr Amara Osei
- Critical — immediate attentionFinding: ventricular tachycardia pattern3 h 41 mOverdue by 101 minutesNobody yet
- Urgent review recommendedSeverity 2 at publication1 h 07 mOpen — 4 hours to targetNobody yet
- Urgent review recommendedInterpretation withheld for signal quality6 h 22 mAcknowledgedDr Wei Lin Tan
An overdue alert means nobody acknowledged it inside this system. This build sends nothing to a pager, an SMS gateway or a mailbox, and says so on its own health screen rather than implying a channel exists.
The audit trail
Opened by: auditor, and the platform operator
The record of who did what, with each entry storing the digest of the entry before it. Deleting or editing a row breaks the chain, so tampering is a detectable gap rather than a quiet success — and that is enforced by a database trigger rather than by the application agreeing to behave. The audit role can read this and cannot touch a clinical record, which is what makes its read independent.
Audit log
Chain head 4b1f9c7a08e3…5d2a · append-only, enforced by the database| At (UTC) | Actor | Event | Entity | Outcome |
|---|---|---|---|---|
| 2026-02-01 09:07:44 | dr.a.osei@hospital.exampleCARDIOLOGIST | ecg.review.recorded | EcgReview 01a0d6cc… | Success |
| 2026-02-01 09:07:51 | dr.a.osei@hospital.exampleCARDIOLOGIST | ecg.signature.created | EcgSignOff 01a0d6cc… | Success |
| 2026-02-01 09:08:02 | dr.a.osei@hospital.exampleCARDIOLOGIST | integration.delivery.completed | IntegrationDelivery 01a0d6ce… | Success |
Each entry stores the digest of the entry before it. Deleting or altering a row breaks the chain, so tampering is a gap that is detectable rather than a change that succeeds — and that is enforced in the database, not by the application agreeing to behave.
The operational figures
Opened by: hospital administrator, auditor
Throughput, turnaround, interpretation coverage and escalation timeliness, each shown with the denominator it was computed against and a sentence saying what it excludes. No diagnostic accuracy figure appears anywhere, because none has been established for this build — a marketing page is the last place to invent one.
Operational analytics
Counted from stored studies · definitions shown with every figureMedian report turnaround
18 m
495 signed studies this period
Receipt to signature. Studies still awaiting signature are excluded, not counted as zero.
Interpretation coverage
94%
1 272 of 1 353 analysed studies
Studies with a published interpretation. Withheld recordings are counted separately below.
Withheld for signal quality
163
12% of studies received
Recordings graded inadequate for automated interpretation. No measurement and no finding exists for these.
Alert acknowledgement inside target
82%
1 186 of 1 446 alerts
Acknowledged before the rule’s target. An overdue alert means nobody acknowledged it inside the system.
Every figure is a count of stored rows and is shown with its denominator. No diagnostic accuracy figure appears anywhere in this system, because none has been established for it.
What a visitor cannot see
Three things the demo does not put on a public page.
Your recordings, not these
Every waveform on this page comes from this system’s own synthetic generator, from the profiles the demonstration deployment serves. An approved workspace reads your own machines’ recordings, arriving by REST or CSV upload with their acquisition parameters intact.
The screens your role opens
The clinical shell is role-aware: a technician does not see a signature button, and an auditor can read the trail without being able to change a clinical record. Each audience page states what that role can do, generated from the same permission table the server enforces.
The parts that are missing
Nothing about a patient is paged, texted or emailed when an alert is raised — a finding reaches the worklist and stops there. Invoices and refunds are not generated, and purchased periods do not renew. The one message this build can send out is operational: that an ECG machine is failing or has stopped sending, to the addresses the deployment names, and only where it has a transport. Each of those is reported by the deployment’s own status endpoint rather than discovered later.
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.