Development and validation of machine learning predictive models for gastric volume based on ultrasonography: A...

Area of Science:

• Anesthesiology
• Medical Imaging
• Machine Learning

Background:

• Aspiration of gastric contents is a critical risk during anesthesia.
• Accurate prediction of gastric volume is essential for patient risk stratification and aspiration prevention.
• Current prediction methods may lack precision.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting gastric volume.
• To compare the performance of ML models against the traditional Perlas model.
• To enhance patient safety by improving gastric volume estimation.

Main Methods:

• A multicentre, cross-sectional study involving adult patients undergoing gastroscopy under intravenous anesthesia.
• Development of eight ML models using features like age, right lateral decubitus (RLD) cross-sectional area (CSA), and Perlas grade, selected via LASSO regression.
• Validation using Bland-Altman analysis and comparison with the Perlas model, including external validation.

Main Results:

• ML models demonstrated superior accuracy in predicting gastric volume compared to the Perlas model (mean bias -0.1 to 2.0 mL vs. 23.5 mL).
• Significant improvement in predicting medium-high gastric volume (AUC: 0.74-0.77 vs. 0.63) and high gastric volume (AUC: 0.85-0.94 vs. 0.74).
• Externally validated models showed high predictive performance (AUCs up to 0.96 for high gastric volume).

Conclusions:

• A novel ML-based predictive model incorporating age, RLD-CSA, and Perlas grade accurately predicts gastric volume.
• The proposed ML model significantly outperforms the traditional Perlas model.
• This advancement offers improved risk stratification and aspiration prevention strategies in anesthesia.