Sparse Decomposition of Heart Rate Using a Bernoulli-Gaussian Model: Application to Sleep Apnoea Detection

Area of Science:

• Cardiology
• Sleep Medicine
• Signal Processing

Background:

• Sleep apnea is a common disorder characterized by repeated breathing interruptions.
• Heart rate variability analysis is crucial for diagnosing sleep-related breathing disorders.
• Existing methods for detecting apnea events from heart rate data have limitations.

Purpose of the Study:

• To develop a sparse decomposition method for analyzing heart rate during sleep.
• To apply this method for the detection of sleep apnea and related respiratory events (apnea-RERA).
• To model heart rate patterns associated with apnea events.

Main Methods:

• Utilizing a sparse decomposition of heart rate signals during sleep.
• Modeling apnea-perturbed heart rate as a Bernoulli-Gaussian (BG) process convolved with a deterministic reference signal.
• Employing deconvolution with sparsity constraints to identify event presence and amplitude.

Main Results:

• The proposed method effectively models tachycardia and bradycardia events following apnea.
• Sparsity imposition allows for a near-syntactic representation of heart rate data.
• Simple detection algorithms applied to this representation show promising results for apnea-RERA detection.

Conclusions:

• Sparse decomposition offers a robust framework for analyzing heart rate during sleep.
• The Bernoulli-Gaussian process model accurately captures heart rate dynamics post-apnea.
• This approach facilitates improved detection of sleep-related respiratory events.