A Deep Learning Method to Predict Difficult Laryngoscopy Using Cervical Spine X-ray Image

Surgery Clinical Trial

Official title:

A Deep Learning Method to Predict Difficult Laryngoscopy Using Cervical Spine X-ray Image: Prospective Validation Study

Clinical Trial Details — Status: Recruiting

Administrative data

• NCT number: NCT05176184
• Other study ID #: 2111-111-1272
• Status: Recruiting
• Start date: December 1, 2021
• Est. completion date: November 25, 2022

Study information

• Verified date: December 2021
• Source: Seoul National University Hospital
• Contact: Hye-yeon Cho, MD
• Phone: +82-10-3808-7110
• Email: bdbd7799[@]gmail.com
• Is FDA regulated: No
• Study type: Observational

Clinical Trial Summary

An unanticipated difficult laryngoscopy is associated with serious airway-related complications. The investigators developed a deep learning-based model that predicts a difficult laryngoscopy (Cormack-Lehane grade 3-4) from a cervical spine lateral X-ray using data from 14,135 patients undergoing thyroid surgery. This model showed excellent predictive performance, which was higher than that of other deep learning architectures. In this study, the investigators prospectively validate the model for predicting a difficult laryngoscopy and compare predictive power with clinical airway evaluation.

Description:

Predicting a difficulty of a laryngoscopy is important for patient safety, as an unanticipated difficult laryngoscopy is associated with serious airway-related complications, such as brain damage, cardiopulmonary arrest, or death. Although clinical predictors, such as the modified Mallampati classification, thyromental distance, inter-incisor gap, and the upper lip bite test, are used for airway evaluation in clinical practice, these indicators have low sensitivity and large inter-assessor variability and require patient cooperation. The investigators developed a deep learning-based model that predicts a difficult laryngoscopy from a cervical spine lateral X-ray using data from 14,135 patients undergoing thyroid surgery. And this study is under submission. This deep learning model showed the highest performance in predicting difficult laryngoscopy compared to other deep learning models (VGG-Net, ResNet, Xception, ResNext, DenseNet, and SENet) with a sensitivity of 95.6%, a specificity of 91.2%, and an area under ROC curve (AUROC) of 0.972. However, as the model was a retrospective design using existing medical records, the presence or absence of cricoid pressure to obtain the optimal laryngoscopy was not evaluated, and not compared with airway evaluations. In this study, the investigators prospectively validate the model for predicting a difficult laryngoscopy and compare predictive power with clinical airway evaluation. If this study prospective confirm our results, this approach can be helpful in improving patient safety and preventing airway-related complications through objective and accurate airway evaluation.

Recruitment information / eligibility

• Status: Recruiting
• Enrollment: 367
• Est. completion date: November 25, 2022
• Est. primary completion date: November 25, 2022
• Accepts healthy volunteers: No
• Gender: All
• Age group: 18 Years and older

Eligibility

Inclusion Criteria:

• elective thyroid surgery under general anesthesia

Exclusion Criteria:

• age < 18 years
• no C-spine lateral X-ray image obtained within 3 months before surgery
• Patient who safety is not guaranteed when using a direct laryngoscope. (poor dental condition, risk of neck extension)
• Patients who not cooperate with the physical examination for airway evaluation

Related Conditions & MeSH terms

• Intubation; Difficult or Failed
• Surgery
• Thyroid

Intervention

Diagnostic Test:
• A deep learning model for predicting a difficult laryngoscopy based on a cervical spine lateral X-ray image
The deep learning model uses the input of preprocessed C-spine lateral X-ray images and outputs the level of difficulty of a laryngoscopy. The easy laryngoscopy is defined as a combination of the Cormack-Lehane grades 1-2 and the difficult laryngoscopy is defined as a combination of grades 3-4. In addition, before general anesthesia, airway evaluations related to the difficulty of laryngoscopy are performed and the results are compared with the actual level of difficulty.

Locations

Country	Name	City	State
Korea, Republic of	Seoul National University Hospital	Seoul	Select A State Or Province

Sponsors (1)

Lead Sponsor	Collaborator
Seoul National University Hospital

Country where clinical trial is conducted

Korea, Republic of, 

References & Publications (3)

Cook TM, MacDougall-Davis SR. Complications and failure of airway management. Br J Anaesth. 2012 Dec;109 Suppl 1:i68-i85. doi: 10.1093/bja/aes393. Review. — View Citation

De Cassai A, Boscolo A, Rose K, Carron M, Navalesi P. Predictive parameters of difficult intubation in thyroid surgery: a meta-analysis. Minerva Anestesiol. 2020 Mar;86(3):317-326. doi: 10.23736/S0375-9393.19.14127-2. Epub 2020 Jan 8. — View Citation

Lundstrøm LH, Vester-Andersen M, Møller AM, Charuluxananan S, L'hermite J, Wetterslev J; Danish Anaesthesia Database. Poor prognostic value of the modified Mallampati score: a meta-analysis involving 177 088 patients. Br J Anaesth. 2011 Nov;107(5):659-67. doi: 10.1093/bja/aer292. Epub 2011 Sep 26. — View Citation

Outcome

Type	Measure	Description	Time frame	Safety issue
Primary	The area under the receiver operating characteristic curve of deep learning model and airway evaluations for predicting a difficult laryngoscopy.	Difficult laryngoscopy definition: Cormack-Lehane grade 3 or 4 . Airway evaluations: Inter-incisor gap (millimeter), thyromental distance (millimeter), thyromental height (millimeter), sternomental distance (millimeter), and modified Mallampati class	during induction of anesthesia
Secondary	The area under the receiver operating characteristic curve of deep learning model and airway evaluations for predicting a difficult intubation.	Difficult intubation: Intubation Difficulty Scale (score)	during induction of anesthesia
Secondary	Other Performances for predicting a difficult laryngoscopy of deep learning model.	sensitivity (percent), specificity(percent), Positive predictive value(percent), Negative predictive value (percent), F1-score, and balanced accuracy.	during induction of anesthesia

External Links

•   DenseNet
•   ResNet
•   ResNext
•   SENet
•   VGG-Net
•   Xception