About TELEPAD & research standards

TELEPAD is built as a research‑grade, prehospital ECG analysis platform. Our algorithms for rhythm classification are evaluated following the ANSI/AAMI EC57 recommended practice for testing and reporting performance of cardiac rhythm algorithms.

Evaluation framework: ANSI/AAMI EC57

ANSI/AAMI EC57:2012/(R)2020 defines how to test and report performance for automated algorithms that detect cardiac rhythm disturbances and measure ST‑segment changes from ECG signals.

What EC57 covers

Beat‑by‑beat performance for QRS, VEB, SVEB detection
Run‑by‑run detection of arrhythmia episodes (AF/VF)
ST‑segment measurement accuracy and reporting
Standardized metrics: sensitivity, positive predictivity, false‑positive rate

How TELEPAD uses EC57

Off‑line validation on MIT-BIH Arrhythmia Database (48 records)
Reporting Se, +P, and false‑positive rates per rhythm class
Record‑by‑record results to highlight clinically significant misses
Separation between algorithm testing (EC57) and clinical performance studies

Model performance (test set)

91.99%

Accuracy

0.9563

ROC-AUC

88.05%

Sensitivity

92.80%

Specificity

Algorithm scope

QRS detection & beat labeling
VEB / SVEB / AF / VF detection
Normal vs Abnormal binary classification

Dataset

MIT-BIH Arrhythmia Database (48 records, 360 Hz)
70% train / 30% test split
~109K beats total after resampling

Planned studies

Retrospective prehospital STEMI detection accuracy
Impact on door-to-balloon and decision time
Prospective pilots in EMS networks

Note on sample rate (360 Hz)

The model is trained on MIT-BIH data recorded at 360 Hz. 180 samples = 0.5 seconds = 1 beat window (90 samples pre-R-peak + 90 post). If your sensor records at a different rate (e.g. ESP32 at 250 Hz or 500 Hz), you must resample to 360 Hz first — otherwise the waveform shape will be distorted and classification accuracy will degrade.

Collaborate on EC57‑aligned validation.

We welcome collaborations with EMS systems, cardiology departments, and academic partners to design EC57‑aligned validation studies and explore new ECG‑based markers for prehospital triage.

Contact research team