Arrhythmias & electrophysiology
Automatic detection of cardiac conditions from photos of electrocardiogram captured by smartphones
Authors' conclusion
Object detection and image segmentation models allow automatic extraction of ECG signals from photos for downstream diagnostics.
Study context
- Population
- ECGs (scanned and photographed) for automatic waveform extraction and AF diagnosis
- Participants
- 40,516 vs 444
- Comparison
- Smartphone-photo ECG waveform extraction pipeline
- Scanned ECGsn = 40,516
- Photographed ECGsn = 444
- Outcome
- Automatic ECG waveform extraction
Performance
As reported in the abstract
| Test or model | Correctly cropped and labelled | Successful voltage-time signal extraction | AF algorithm sensitivity | AF algorithm specificity | AF algorithm PPV | AF algorithm NPV | AF algorithm F1 score |
|---|---|---|---|---|---|---|---|
| Scanned | 12 828 of 13 258 (96.8%) | 11 604 of 12 735 (91.1%) | |||||
| Photographed | 5399 of 5743 (94.0%) | 5062 of 5752 (88.0%) | 91.3% | 94.2% | 95.6% | 88.6% | 93.4% |
Correctly cropped and labelled
Scanned12 828 of 13 258 (96.8%)
Photographed5399 of 5743 (94.0%)
Successful voltage-time signal extraction
Scanned11 604 of 12 735 (91.1%)
Photographed5062 of 5752 (88.0%)
AF algorithm sensitivity
Scanned— not stated in the abstract
Photographed91.3%
AF algorithm specificity
Scanned— not stated in the abstract
Photographed94.2%
AF algorithm PPV
Scanned— not stated in the abstract
Photographed95.6%
AF algorithm NPV
Scanned— not stated in the abstract
Photographed88.6%
AF algorithm F1 score
Scanned— not stated in the abstract
Photographed93.4%
Original publication
Automatic detection of cardiac conditions from photos of electrocardiogram captured by smartphones. Heart. Published 20 May 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →