Respiratory failure & ventilation
A Comparison of Five Algorithmic Methods and Machine Learning Pattern Recognition for Artifact Detection in Electronic Records of Five Different Vital Signs: A Retrospective Analysis
Authors' conclusion
No single artifact detection method consistently performed well across different vital signs and clinical settings.
Study context
- Population
- Intraoperative OR and intensive care unit (ICU) electronic health record datasets (heart rate, oxygen saturation, BP, temperature, capnometry)53 OR and 53 ICU
- Participants
- 106 participants
- Comparison
- Five classical artifact detection algorithms and an LSTM neural network vs human expert annotation
- Outcome
- Sensitivity and specificity of each artifact detection algorithm vs human expert reference
Methodological detailssingle-centre
- Single-centre
Performance
As reported in the abstract
Artifact detection performance
| Test or model | Sensitivity, heart rate (ICU) | Sensitivity, systolic invasive BP (OR) | Sensitivity, systolic invasive BP (ICU) | Sensitivity, temperature (OR) | Sensitivity, oxygen saturation (OR) |
|---|---|---|---|---|---|
| Neural network | 33.6% (95% CI, 33.1 to 44.6) | 62.2% (95% CI, 57.5 to 71.9) | 60.7% (95% CI, 57.3 to 71.8) | 76.1% (95% CI, 63.6 to 89.7) | — not stated in the abstract |
| Z-value | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract | 88.9% |
Sensitivity, heart rate (ICU)
Neural network33.6% (95% CI, 33.1 to 44.6)
Z-value— not stated in the abstract
Sensitivity, systolic invasive BP (OR)
Neural network62.2% (95% CI, 57.5 to 71.9)
Z-value— not stated in the abstract
Sensitivity, systolic invasive BP (ICU)
Neural network60.7% (95% CI, 57.3 to 71.8)
Z-value— not stated in the abstract
Sensitivity, temperature (OR)
Neural network76.1% (95% CI, 63.6 to 89.7)
Z-value— not stated in the abstract
Sensitivity, oxygen saturation (OR)
Neural network— not stated in the abstract
Z-value88.9%
— not stated in the abstract
- Data points annotated as artifacts by human experts
- 5,167 (1.3%)
Original publication
A Comparison of Five Algorithmic Methods and Machine Learning Pattern Recognition for Artifact Detection in Electronic Records of Five Different Vital Signs: A Retrospective Analysis. Anesthesiology. Published 11 March 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →