Deep‐learning model associating lateral cervical radiographic features with Cormack–Lehane grade 3 or 4 glottic view
Authors' conclusion
Radiographic features around the hyoid bone, pharynx and cervical spine were associated with grade 3 and 4 glottic views.
Study context
- Population
- Thyroid surgery patients; prediction of Cormack-Lehane grade 3 or 4 direct laryngoscopy views253 (4%) grade 3 or 4
- Participants
- 5,939 patients
- Comparison
- New convolutional neural network model vs six similar models
- Outcome
- Brier score for prediction of Cormack-Lehane grade 3 or 4 glottic view
Methodological detailsdeep-learning model development, single-centre, 10 randomly sampled datasets used to train the model
- Deep-learning model development
- Single-centre
- 10 randomly sampled datasets used to train the model
Performance
As reported in the abstract
Brier score (95%CI; lower = better)
New model
0.023
95% CI 0.021–0.025
VGG
0.034
95% CI 0.034–0.035
ResNet
0.033
95% CI 0.033–0.035
Xception
0.032
95% CI 0.031–0.033
ResNext
0.033
95% CI 0.032–0.033
DenseNet
0.030
95% CI 0.029–0.032
SENet
0.031
95% CI 0.029–0.032
- New model lower than each other model
- all p < 0.001
Further findings
| Outcome | Values |
|---|---|
| Performance of the new model | New model R2 0.428 (0.388-0.468) · mean squared error 0.023 (0.021-0.025) · mean absolute error 0.048 (0.046-0.049) · balanced accuracy 0.713 (0.684-0.742) · AUC 0.965 (0.962-0.969) |
- Performance of the new model: mean (95%CI)
Original publication
Deep‐learning model associating lateral cervical radiographic features with Cormack–Lehane grade 3 or 4 glottic view. Anaesthesia. Published 5 October 2022. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →