Respiratory failure & ventilation
A deep learning model enables accurate prediction and quantification of pulmonary edema from chest X-rays
Authors' conclusion
Deep learning can quantify pulmonary edema as measured by EVLWI with high accuracy.
Study context
- Population
- ICU patients with chest radiography and transpulmonary thermodilution (TPTD) measurement within 24 h471 X-rays
- Participants
- 431 patients
- Comparison
- Binary model vs Multiclass models
- Outcome
- Prediction of pulmonary edema severity (EVLWI from TPTD) from chest X-rays
Methodological detailsdeep learning, single-centre ICU
- Deep learning
- Single-centre ICU
Performance
As reported in the abstract
Classification performance
| Test or model | Accuracy | AUROC | Mathews correlation coefficient (MCC) | Accuracy (range) | AUROC (range) | MCC (range) |
|---|---|---|---|---|---|---|
| Binary model | 0.93 | 0.98 | 0.86 | |||
| Multiclass models | 0.90 and 0.95 | 0.97 and 0.99 | 0.86 and 0.92 |
Accuracy
Binary model0.93
Multiclass models— not stated in the abstract
AUROC
Binary model0.98
Multiclass models— not stated in the abstract
Mathews correlation coefficient (MCC)
Binary model0.86
Multiclass models— not stated in the abstract
Accuracy (range)
Binary model— not stated in the abstract
Multiclass models0.90 and 0.95
AUROC (range)
Binary model— not stated in the abstract
Multiclass models0.97 and 0.99
MCC (range)
Binary model— not stated in the abstract
Multiclass models0.86 and 0.92
Original publication
A deep learning model enables accurate prediction and quantification of pulmonary edema from chest X-rays. Crit Care. Published 26 May 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →