Obstructive sleep apnea syndrome detection based on ballistocardiogram via machine learning approach

Area of Science:

• Biomedical Engineering
• Respiratory Medicine
• Sleep Science

Background:

• Obstructive sleep apnea (OSA) is a prevalent respiratory disorder, particularly affecting the elderly.
• Traditional polysomnography (PSG) for OSA detection can be intrusive and prone to sensor dislodgement.
• There is a need for comfortable, accurate, and non-disruptive OSA monitoring methods.

Purpose of the Study:

• To develop and validate a non-contact method for detecting obstructive sleep apnea (OSA).
• To assess the efficacy of using piezoelectric sensors in a mattress for sleep apnea detection.
• To improve patient comfort and compliance in OSA diagnosis.

Main Methods:

• Utilized piezoelectric ceramic sensors embedded in a mattress to capture chest and abdominal pressure variations.
• Extracted heart rate and respiratory rate from processed pressure signals.
• Calculated Heart Rate Variability (HRV) from heartbeat intervals.
• Employed feature extraction from heartbeat and respiratory signals over time intervals.
• Implemented a classification model with model fusion technology for OSA prediction.

Main Results:

• The non-contact mattress system successfully captured physiological signals indicative of sleep.
• Extracted heart rate, respiratory rate, and HRV provided data for sleep apnea analysis.
• The classification model demonstrated effectiveness in predicting sleep apnea events.
• Model fusion enhanced the overall accuracy of OSA detection.

Conclusions:

• The proposed non-contact mattress-based system offers an effective and comfortable alternative for obstructive sleep apnea detection.
• This technology minimizes sleep disturbance and sensor-related issues associated with traditional PSG.
• Further research may validate this method for widespread clinical use in OSA screening and monitoring.