講演抄録/キーワード |
講演名 |
2019-11-15 11:25
Sleep Apnea Detection Using Deep Convolutional Neural Network ○Hnin Thiri Chaw・Sinchai Kamolphiwong・Krongthong Wongsritrang(PSU) IA2019-39 |
抄録 |
(和) |
Sleep apnea is the breathing disorder in which the cessation of the air flow occurs while sleeping. Deep convolutional neural network is good at effective and powerful feature learning ability in order to detect sleep apnea. Moreover, it is also suitable for large and complex nature of the datasets such as the nature of polysomnography signals, EEG(electroencephalography) signal, ECG(electrocardiography) signal, oximetry signal. The nature of the respiratory signals is also very complex and it contain nasal pressure signal, thermistor signal, oximetry signal, thoracic signal and abdominal signal. In this paper, the detection of the sleep apnea using deep convolutional network will be used and the signal that is used in this paper is oximetry signal which is also known as SPO2 signal. In addition, the model that is used in this study is known as type 4 sleep studies which focus more on portability and reduction of the signals in order to cover the limitations of the type 1 sleep studies which require a number of signals known as polysomnography. The total number of SPO2 samples from 50 patients is 190,000 samples and subject specific scenario validation is used with a split rate of 0.2 in each empirical study. The performance of the overall accuracy of sleep apnea detection is 91.3085% with the loss rate the prediction of 2.3 using cross entropy cost function. |
(英) |
Sleep apnea is the breathing disorder in which the cessation of the air flow occurs while sleeping. Deep convolutional neural network is good at effective and powerful feature learning ability in order to detect sleep apnea. Moreover, it is also suitable for large and complex nature of the datasets such as the nature of polysomnography signals, EEG(electroencephalography) signal, ECG(electrocardiography) signal, oximetry signal. The nature of the respiratory signals is also very complex and it contain nasal pressure signal, thermistor signal, oximetry signal, thoracic signal and abdominal signal. In this paper, the detection of the sleep apnea using deep convolutional network will be used and the signal that is used in this paper is oximetry signal which is also known as SPO2 signal. In addition, the model that is used in this study is known as type 4 sleep studies which focus more on portability and reduction of the signals in order to cover the limitations of the type 1 sleep studies which require a number of signals known as polysomnography. The total number of SPO2 samples from 50 patients is 190,000 samples and subject specific scenario validation is used with a split rate of 0.2 in each empirical study. The performance of the overall accuracy of sleep apnea detection is 91.3085% with the loss rate the prediction of 2.3 using cross entropy cost function. |
キーワード |
(和) |
deep convolutonal neural network / sleep apnea detection / type 4 sleep studies / polysomnography / spo2 / oximetry / deep learning / sleep apnea prediction |
(英) |
deep convolutonal neural network / sleep apnea detection / type 4 sleep studies / polysomnography / spo2 / oximetry / deep learning / sleep apnea prediction |
文献情報 |
信学技報, vol. 119, no. 291, IA2019-39, pp. 67-71, 2019年11月. |
資料番号 |
IA2019-39 |
発行日 |
2019-11-07 (IA) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
IA2019-39 |
|