Obstructive sleep apnea syndrome detection based on ballistocardiogram via machine learning approach.
Journal:
Mathematical biosciences and engineering : MBE
Published Date:
Jun 19, 2019
Abstract
Obstructive sleep apnea (OSA) is a common sleep-related respiratory disease that affects people's health, especially in the elderly. In the traditional PSG-based OSA detection, people's sleep may be disturbed, meanwhile the electrode slices are easily to fall off. In this paper, we study a sleep apnea detection method based on non-contact mattress, which can detect OSA accurately without disturbing sleep. Piezoelectric ceramics sensors are used to capture pressure changes in the chest and abdomen of the human body. Then heart rate and respiratory rate are extracted from impulse waveforms and respiratory waveforms that converted by filtering and processing of the pressure signals. Finally, the Heart Rate Variability (HRV) is obtained by processing the obtained heartbeat signals. The features of the heartbeat interval signal and the respiratory signal are extracted over a fixed length of time, wherein a classification model is used to predict whether sleep apnea will occur during this time interval. Model fusion technology is adopted to improve the detection accuracy of sleep apnea. Results show that the proposed algorithm can be used as an effective method to detect OSA.
Authors
Keywords
Algorithms
Ballistocardiography
Decision Trees
Diagnosis, Computer-Assisted
Electrocardiography
False Positive Reactions
Heart Rate
Humans
Machine Learning
Nonlinear Dynamics
Reproducibility of Results
Risk
Sensitivity and Specificity
Signal Processing, Computer-Assisted
Sleep
Sleep Apnea, Obstructive