Presentation (CTI)

OBJECTIVES: This study introduces a deep learning-based system for automating the assessment of tracheal intubation procedures by resident physicians.
METHODS: The study includes resident physicians involved in performing tracheal intubation procedures, and the data is collected from the medical training center. The experimental group employs a deep learning-based system for the assessment of tracheal intubation procedures, whereas the control group undergoes evaluation by professional human assessors. Comparative analysis of the two groups is conducted using T-test for inter-group comparisons and C2-test for inter-group comparisons of count data. Incremental cost-effectiveness ratio analysis is employed to evaluate the effectiveness of the management approach.
RESULTS: The control group comprises 103 resident physicians participating in the training, while the experimental group consists of 121 participants. There are no significant differences in age and gender between the two groups. Following the introduction of the deep learning automated assessment system, the success rate of tracheal intubation significantly increased (81.9% vs 75.2%, p=0.019). The average training time for physicians in the experimental group (8.65 mins vs 12.00 mins, p>0.05) decreased by 3.35 minutes compared to the control group. Thus, the implementation of this intelligent system not only reduces manpower costs but also enhances training effectiveness, making it cost-effective.
CONCLUSIONS: Establishing a deep learning-based intelligent system for the automated assessment of tracheal intubation procedures can effectively improve physicians' learning efficiency and, in the long term, reduce healthcare costs. However, to ensure the accuracy and reliability of results, future research should involve larger-scale, multicenter randomized controlled trials to validate the practicality and economic benefits of this algorithm.