Pharyngeal wall vibration detection using an artificial neural network

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Sleep Medicine

Background:

• Obstructive sleep apnea (OSA) affects numerous adults, necessitating effective therapeutic interventions.
• Continuous positive airway pressure (CPAP) is a primary treatment, but its efficacy can be improved with personalized pressure adjustments.
• Pharyngeal wall vibration (PWV) is a physiological signal that precedes OSA events, offering a potential target for automated therapy modulation.

Purpose of the Study:

• To develop and evaluate an artificial neural network (ANN) for automated detection of pharyngeal wall vibrations (PWV).
• To assess the potential of ANN-based PWV detection in enhancing automatic positive airway pressure (APAP) therapy for OSA patients.

Main Methods:

• An artificial neural network with 15 inputs, one output, and two hidden layers (each with two Adaline-nodes) was designed for PWV detection.
• The ANN was trained using nasal mask pressure data from five diagnosed OSA patients.
• The detector's performance was validated on data from five independent OSA patients.

Main Results:

• The ANN-based detector achieved an average accuracy of approximately 92% in identifying PWV events.
• The system demonstrated a high accuracy of approximately 98% in correctly distinguishing normal breathing patterns.
• The detector's accuracy remained consistent regardless of the therapeutic pressure levels used.

Conclusions:

• An ANN-based system can reliably detect pharyngeal wall vibrations (PWV), a precursor to obstructive sleep apnea (OSA) events.
• This automated detection method holds significant promise for advancing continuous positive airway pressure (CPAP) therapy towards more personalized automatic positive airway pressure (APAP) systems.
• The robustness of the ANN detector across different pressure levels suggests its practical applicability in real-world sleep apnea management.