Deep Learning-Based Localization and Detection of Malpositioned Endotracheal Tube on Portable Supine Chest Radiographs in Intensive and Emergency Medicine: A Multicenter Retrospective Study*
Authors' conclusion
The derived CAD system could localize ETT and detect ETT malposition with excellent performance, especially for endobronchial intubation, and with favorable potential for external generalizability.
Study context
- Population
- Portable supine chest radiographs of patients aged ≥20 years examined in EDs or ICUs; training set NTUH-1519 (National Taiwan University Hospital (NTUH), 2015–2019)5,767 training images
- Participants
- 955 vs 656
- Comparison
- NTUH-20 vs NTUH-YB
- NTUH-20n = 955
- NTUH-YBn = 656
- Outcome
- ETT localization and detection
Methodological detailsmulticentre, chronological and geographical external test sets, DeepLabv3+ with ResNeSt50 backbone for segmentation; DenseNet121 for classification, All CXRs annotated with pixel-level ETT labels and image-level labels of ETT presence and malposition
- Multicentre
- Chronological and geographical external test sets
- DeepLabv3+ with ResNeSt50 backbone for segmentation; DenseNet121 for classification
- All CXRs annotated with pixel-level ETT labels and image-level labels of ETT presence and malposition
Performance
As reported in the abstract
| Test or model | Dice coefficient (segmentation) | AUC, ETT presence | AUC, ETT malposition | AUC, endobronchial intubation |
|---|---|---|---|---|
| NTUH-20 | 0.854 (95% CI, 0.824-0.881) | 1.000 (95% CI, 0.999-1.000) | 0.847 (95% CI, 0.671-0.980) | 0.991 (95% CI, 0.969-1.000) |
| NTUH-YB | 0.839 (95% CI, 0.820-0.857) | 0.994 (95% CI, 0.984-1.000) | 0.734 (95% CI, 0.630-0.833) | 0.966 (95% CI, 0.933-0.991) |
Dice coefficient (segmentation)
NTUH-200.854 (95% CI, 0.824-0.881)
NTUH-YB0.839 (95% CI, 0.820-0.857)
AUC, ETT presence
NTUH-201.000 (95% CI, 0.999-1.000)
NTUH-YB0.994 (95% CI, 0.984-1.000)
AUC, ETT malposition
NTUH-200.847 (95% CI, 0.671-0.980)
NTUH-YB0.734 (95% CI, 0.630-0.833)
AUC, endobronchial intubation
NTUH-200.991 (95% CI, 0.969-1.000)
NTUH-YB0.966 (95% CI, 0.933-0.991)
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Deep Learning-Based Localization and Detection of Malpositioned Endotracheal Tube on Portable Supine Chest Radiographs in Intensive and Emergency Medicine: A Multicenter Retrospective Study*. Crit Care Med. Published 14 December 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →