Model-Based Analysis of Pulse Transit Time-Derived Features for the Classification of Obstructive and Central Apneas.

Area of Science:

• Cardiovascular Physiology
• Sleep Medicine
• Biomedical Engineering

Background:

• Pulse transit time (PTT) is a non-invasive method for assessing cardiovascular function.
• Sleep apnea syndrome (SAS) diagnosis can benefit from objective, automated tools.
• Current PTT applications for SAS lack standardized, interpretable features.

Purpose of the Study:

• To develop and validate novel PTT-derived features for classifying obstructive and central apneas in SAS.
• To establish a physiologically explainable model for PTT variations during apneas.
• To improve automatic detection and characterization of SAS subtypes.

Main Methods:

• Extracted features from PTT and its oscillations (ΔPTT) in 26 HYPNOS study patients.
• Trained a random forest classifier using these features to rank their importance.
• Employed Morris sensitivity analysis on an integrated cardio-respiratory model simulating PTT.

Main Results:

• Achieved an Area Under the Curve (AUC) of 0.83 ± 0.12 for apnea classification.
• Identified features related to ΔPTT changes during apnea as most significant.
• Demonstrated feature sensitivity to parameters influencing blood CO2 levels.

Conclusions:

• Proposed a novel integrated model for PTT in sleep apnea.
• Enabled physiological interpretation of informative PTT features.
• Facilitated improved classification of obstructive and central apneas using PTT.