Development of Algorithm for Autotitrating CPAP for Treating Obstructive Sleep Apnea
Date Issued
2016
Date
2016
Author(s)
Chiao, Yu-An
Abstract
For patients suffering from sleep apnea, continuous positive airway pressure (CPAP) is the most recommended therapy currently. However, manual CPAP titration at sleep center is time consumption and high-cost. Therefore, PAP machine (APAP) which can detect the breathing event and further automatically adjust the therapeutic pressure has been demonstrated to lower therapeutic pressure than fixed-pressure CPAP. Also, patients prefer APAP than fixed-pressure CPAP though the compliance was similar between two devices. In this thesis, an automated CPAP titration algorithm is proposed. We verified our algorithm with database of overnight CPAP titration in the sleep center of NTUH. This novel algorithm only used PAP flow signal. Apnea and hypopnea detection can be realized by signal processing of several signals. Besides, we extracted several features from PAP flow signal, do feature selection, put them into deep-learning neural network to generate a classifier for snore detection, and use cross-validation method to do verification. A simple recheck method was also introduced to do CSA detection. Finally, the therapeutic pressure of CPAP was determined with algorithm according to the aforementioned event detections.
Subjects
sleep apnea
automatically-adjusting positive airway pressure
breathing event detection
neural network
therapeutic pressure autotitration
Type
thesis