ecg/12-lead

What it does

Resting 12-Lead ECG reads one standard 10-second ECG, whether it arrives as a digital file from the electrocardiograph, a PDF export, or a phone photo or scan of the paper printout. Images are first digitised into calibrated lead signals with a confidence per lead, so a photographed ECG is measured the same way as a native one.

The read returns rhythm and heart rate, PR, QRS, QT and QTc with axes, the full statement list mapped to AHA/ACCF/HRS codes, and critical flags: STEMI and STEMI-equivalents, ventricular tachycardia, complete heart block, QTc of 500 ms or more, pre-excitation, Brugada type 1 and a hyperkalaemia pattern. AI-ECG screens return the probability of reduced ejection fraction and of structural heart disease, the future risk of atrial fibrillation and a potassium estimate.

Intended use

Decision support for the physician reading an ECG alone: the night-shift resident asking whether this is an occlusion MI, the general practitioner checking a QTc before a QT-prolonging drug, the emergency doctor looking for a dangerous rhythm or a lead reversal. Critical flags show the leads and beats that triggered them.

Who it is for

Emergency physicians, general practitioners, internists, cardiology residents and cardiologists. Developers use the same run to add ECG interpretation, digitisation of paper archives or wave delineation to their own systems.

Accepted input

• DICOM waveforms: 12-lead ECG, General ECG and General 32-bit ECG Waveform objects; Encapsulated PDF ECG reports.
• Signal files: SCP-ECG (EN 1064 / ISO 11073-91064), HL7 v3 annotated ECG (aECG) XML, vendor XML exports, WFDB and EDF/EDF+.
• Images: vector or raster PDF, and JPEG, PNG or HEIC photos and scans in 3×4 with rhythm strip, 6×2 or 12×1 layouts. Vector PDFs are read from their drawing paths rather than traced from pixels.

The device's own printed interpretation on a PDF or photo is masked before the trace is read, so it cannot anchor the result.

Requirements

Requirement	Why
Sampling rate of at least 250 Hz (500 Hz preferred)	ST and interval measurement need fine time resolution
Amplitude scale and lead labels in signal files	Millivolt values and lead-specific criteria
On images: all 12 leads and the calibration pulse visible	Paper speed (25/50 mm/s) and gain (5/10/20 mm/mV) come from the pulse and grid when not printed
Patient age and sex	Normal ranges, QTc thresholds and paediatric norms

Optional context

Symptoms and onset time, serum potassium, QT-prolonging drugs and a prior ECG (in prior_files) sharpen the guidance and enable the serial comparison.

The result

Section	Content
Digitisation	For images: 12 calibrated lead signals with per-lead confidence, layout, paper speed and gain
Recording quality	Noise, baseline wander, saturation, missing leads, suspected limb or precordial lead reversal, implausible gain
Rhythm and conduction	Rhythm class, ectopy, AV block degree, bundle-branch and fascicular blocks, pre-excitation, with evidence beats
Intervals and axes	HR, PR, QRS, QT, QTc (Bazett and Fridericia), P/QRS/T axes, J-point and ST amplitude per lead, median beats
STEMI and occlusion MI	ST criteria per lead, STEMI-equivalents (posterior, de Winter, Sgarbossa in LBBB or pacing), occlusion-MI probability, likely culprit territory
Comparison with prior ECG	New or resolved findings: new LBBB, new Q waves, dynamic ST-T change, QTc change
Low EF screen	Probability of LVEF ≤ 40–45 %, with echocardiography as the next test when positive
Structural heart disease screen	Estimated probability of moderate or greater structural heart disease, with components for valve disease, wall thickness ≥ 13 mm, RV dysfunction, pericardial effusion and PASP ≥ 45 mmHg
Hyperkalaemia estimate	Potassium band and probability of K+ ≥ 6.0 mmol/L, with the ECG pattern evidence
Future AF risk	1–5-year risk of incident AF from a sinus-rhythm ECG
Wave delineation	Per-beat P, QRS and T onsets, peaks and offsets with beat labels
Draft report	Opt-in (options.report): an English narrative that interprets the result for the reader (findings, impression, limitations); every number is checked against the findings

Standards

• DICOM: Comprehensive SR TID 3700 ECG Report (global measurements TID 3713, lead measurements TID 3714, qualitative analysis TID 3717, summary TID 3719, prior ECG TID 3702); Waveform Annotation SR for fiducials; digitised signals re-encoded as 12-lead ECG Waveform marked as derived; Encapsulated PDF.
• Interchange: HL7 aECG and SCP-ECG export; FHIR R4 Observation (LOINC 11524-6 with heart rate, PR, QRS, QT, QTc and axis components) and DiagnosticReport (LOINC 18844-1).
• Statements: AHA/ACCF/HRS 2007 Part II diagnostic statement list, with SCP-ECG and SNOMED CT cross-codes.
• Guidance: Fourth Universal Definition of MI (2018); 2023 ESC and 2025 ACC/AHA/ACEP/NAEMSP/SCAI ACS guidelines; AHA/ACCF/HRS 2009 Parts IV and VI; AHA/ACCF 2010 statement on torsade de pointes; 2024 ESC and 2023 ACC/AHA/ACCP/HRS atrial fibrillation guidelines; 2018 ACC/AHA/HRS bradycardia guideline; 2022 ESC ventricular arrhythmia guidelines; 2017 international criteria for ECG interpretation in athletes.

QT notes

QTc is reported by Bazett and Fridericia, with Fridericia preferred at high heart rates. A QTc of 500 ms or more, or a rise of 60 ms or more against the prior ECG, is flagged with the AHA/ACCF 2010 advice to review QT-prolonging drugs.

Reading limits

Posterior infarction shows as ST depression in V1–V3; leads V7–V9, which show it directly, are not part of a standard 12-lead. ST changes in LBBB and paced rhythm are read with Sgarbossa criteria. Paediatric ECGs use age-specific norms. A photographed ECG carries a lower per-lead confidence than a native file, and the result shows it.