Evaluation of Double Lumen Tube Intubation Difficulty With Photo-Based Artificial Intelligence Algorithms

Efficacy and Reliability of Photo-Based Artificial Intelligence Algorithms in Assessing Difficulty of Intubation With a Double-Lumen Tube

The complexity and difficulty of intubation with double lumen tubes requires the use of advanced technologies in the management of this procedure. The potential of photo-based artificial intelligence algorithms to predict and minimize the difficulties encountered during intubation is the main motivation for this study.

The utilization of artificial intelligence algorithms within the domain of airway management holds considerable promise in providing real-time feedback to anesthesiologists, enhancing the efficacy of intubation procedures, and reducing the occurrence of complications. Specifically, photo-based AI systems can facilitate a more comprehensive understanding of airway anatomy by analyzing images captured prior to and during intubation, thereby enhancing the management of complex cases.The objective of this study is to examine the efficacy and reliability of photo-based artificial intelligence algorithms in evaluating the complexity of intubation with a double lumen tube.The integration of artificial intelligence into the intubation process is intended to enhance patient outcomes and establish a new benchmark for anesthesia practice. This study aims to address the existing gap in the literature and provide innovative approaches to clinical practice.

Informed consent was obtained from patients undergoing thoracic surgery operations, and demographic data (age, height, body weight, body mass index, gender), American Society of Anesthesiologists (ASA) score, type of operation, and comorbid diseases (diabetes mellitus, hypertension, coronary artery disease, chronic kidney disease, chronic obstructive pulmonary disease, asthma, obstructive sleep apnea) were obtained. Thoracic and/or extrath oracic malignancy history), parameters considered as risk factors for difficult intubation (history of previous difficult intubation, LEMON criteria (look externally, evaluate, mallampathy, obstruction, neck mobility), upper lip bite test) and photographs of the patients (including head and neck region) will be recorded in six different directions and ways with a professional camera (actively used in our hospital) in the preoperative period. During the intraoperative phase, the Cormack-Lehane scoring system will be employed, and the intubation process with a double-lumen tube will be evaluated for ease or difficulty. Intraoperative complications related to the operation will also be documented.The data will then be processed using Python 3 programming language and open-source libraries to calculate artificial intelligence algorithms. In the event of incomplete patient data, data imputation techniques will be employed to supplement the artificial intelligence program.

The primary outcome variable of the study is the rate at which the photo-based artificial intelligence algorithm predicts whether intubation with a double lumen tube is easy or difficult.The secondary outcome variable is the comparison of the rate of prediction of intubation with double lumen tube by photo-based artificial intelligence algorithms and the rate of prediction of intubation with double lumen tube by conventional methods.

Study Overview

• Status: Completed
• Conditions: Intubation; Difficult; Double Lumen Tube Intubation; Artificial Intelligence (AI)
• Intervention / Treatment: Diagnostic test: Intubation Difficulty Scale; Other: Artificial Intelligence

• Study Type: Observational
• Enrollment (Actual): 260

Contacts and Locations

Study Locations

• Turkey (Türkiye)
  • Keçiören
    • Ankara, Keçiören, Turkey (Türkiye), 06290
      • Ankara Atatürk Sanatoryum Training and Research Hospital

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

The sample size was calculated as 214 patients with a power of 0.90, 5% type-1 error, 81.8% sensitivity and 26.7% prevalence (AUC=0.864) with a power of 0.90, 5% type-1 error, 81.8% sensitivity and (AUC=0.864) 26.7% prevalence in the Sample Size Estimation for Diagnostic Accuracy Studies calculated for the study considering the literature data. Since 214 patients would be used to teach the AI easy and difficult intubation and 46 patients would be used to test the AI, a total of 260 patients were recruited.

Description

Inclusion Criteria:

• Undergoing thoracic surgery
• Giving informed consent
• Over 18 years of age
• Double lumen tube used for intubation
• ASA (American Society of Anesthesiologist)1-2-3

Exclusion Criteria:

• Emergency surgeries
• ASA 4 and above
• Head and neck tumor, history of surgery/RT related to tumor
• Presence of syndrome that will cause difficult intubation

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Intubation - Difficult; According to the Intubation Difficulty Scale (IDS), a score of > 5 was defined as difficult intubation.	Diagnostic test: Intubation Difficulty Scale; The Intubation Difficulty Scale (IDS) is an objective way to classify easy and difficult intubation. A score ≤ 5 indicates an easy or mildly difficult intubation, while IDS > 5 suggests difficult intubation, requiring additional techniques or attempts.; Other: Artificial Intelligence; The program made with Python 3 programming language using open source libraries. It will be developed to predict difficult intubation with 6 different photo data of patients, this process will be taught with a learning process and then tested.
Intubation - Easy; According to the Intubation Difficulty Scale (IDS), a score of ≤ 5 was defined as easy intubation.	Diagnostic test: Intubation Difficulty Scale; The Intubation Difficulty Scale (IDS) is an objective way to classify easy and difficult intubation. A score ≤ 5 indicates an easy or mildly difficult intubation, while IDS > 5 suggests difficult intubation, requiring additional techniques or attempts.; Other: Artificial Intelligence; The program made with Python 3 programming language using open source libraries. It will be developed to predict difficult intubation with 6 different photo data of patients, this process will be taught with a learning process and then tested.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Intubation Difficulty Scale (IDS)	The Intubation Difficulty Scale (IDS) is an objective way to classify easy and difficult intubation. A score ≤ 5 indicates an easy or mildly difficult intubation, while IDS > 5 suggests difficult intubation, requiring additional techniques or attempts.	During the operation

Collaborators and Investigators

• Sponsor: Ankara Ataturk Sanatorium Training and Research Hospital
• Investigators: Principal Investigator: Onur Küçük, Specialist, Ankara Atatürk Sanatoryum Training and Research Hospital

Study record dates

Study Major Dates

• Study Start (Actual): December 1, 2024
• Primary Completion (Actual): March 1, 2025
• Study Completion (Actual): March 30, 2025

Study Registration Dates

• First Submitted: February 17, 2025
• First Submitted That Met QC Criteria: February 17, 2025
• First Posted (Actual): February 21, 2025

Study Record Updates

• Last Update Posted (Estimated): September 8, 2025
• Last Update Submitted That Met QC Criteria: September 4, 2025
• Last Verified: February 1, 2025

More Information

Terms related to this study

• Keywords: artificial intelligence; thoracic surgery; airway management; Difficult intubation; Double lumen tube
• Additional Relevant MeSH Terms: Algorithms; Mathematical Concepts; Artificial Intelligence
• Other Study ID Numbers: 2024-KAEK-24/07

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No
• product manufactured in and exported from the U.S.: No