Predicting and Understanding Difficult Mask Ventilation: Classification and Generative Networks

Mc Auley, Mícáel Tomas (2025) Predicting and Understanding Difficult Mask Ventilation: Classification and Generative Networks. Master's Thesis / Essay, Computational Cognitive Science.

Preview	Text CCS2025McAuleyM.pdf Download (2MB) | Preview
	Text Toestemming.pdf Restricted to Registered users only Download (268kB)

Abstract

Mask ventilation is a vital aspect of airway management in clinical, emergency, and surgical settings, yet unexpected difficulty in mask ventilation remains a significant cause of morbidity and mortality. Current predictive methods for identifying at-risk patients are often insufficient. Recent research suggests that vocal acoustics, influenced by upper airway anatomy, may offer a novel predictive tool. This study explores human-centered artificial intelligence techniques, namely Convolutional Neural Networks and Variational Autoencoders, to improve the prediction and understanding of difficult mask ventilation. By applying AI-driven analysis to vocal biomarkers, this study aims to convert anecdotal intuition-based observations of voice and airway difficulty into an evidence-based predictive model. The first objective involves developing a predictive model using mel-frequency spectrograms, a human-interpretable audio representation, to classify patients based on vocal patterns. The second objective investigates the generation of synthetic voice samples simulating both high-risk and routine patients, enhancing model interpretability and serving as a potential educational tool. Moderate success in classification was achieved, with a model obtaining a slight but meaningful ability to distinguish difficult mask ventilation from a patients voice alone. The generative model’s computational requirements exceeded available resources, preventing successful synthesis of voice samples.

Item Type:	Thesis (Master's Thesis / Essay)
Supervisor name:	Schippers, M.B.
Degree programme:	Computational Cognitive Science
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	17 Jun 2025 13:25
Last Modified:	17 Jun 2025 13:25
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/35377

Actions (login required)

View Item