Return to search

AHI prediction improvement by oxyhemoglobin desaturation features with new baseline definition and EEG wake information

The diagnosis of obstructive sleep apnea (OSA) syndrome is overnight PSG (mutli-channel system). But it¡¦s hard to be popularized for the general population (about twenty channel signals). In recent decades, several researches were devoted to a replacement system with only one channel signal (oxyhemoglobin saturation). However, it¡¦s hard to match PSG system¡¦s report without EEG wake information. Consequently, two channels (oxyhemoglobin saturation and EEG) were used of this study to enhance the AHI (estimation index for sleep apnea) prediction performance. After surveying the most recent studies, this work proposes a new basleline removal technique for oxygen saturation signal (SpO2) by using median filter. It was proved this technique improves the diagnostic accuracy for OSA. Furthermore, it is also found that by removing the wake periods, diagnostic accuracy can be improved further.
By counting the number of times that the desaturation level has dropped more than 2% for at least 3 seconds, the correlation coefficient between AHI and proposed feature is 0.9218. In addition, by removing the wake period, this correlation increases to 0.9425. By using this feature to classify patients with AHI value larger than 5, the proposed approach achieves 93.78% accuracy, 95.94% sensitivity, 78.87% specificity f. Such results demonstrate the feasibility of using single SpO2 channel system for OSA diagnosis.

Identiferoai:union.ndltd.org:NSYSU/oai:NSYSU:etd-0717109-161820
Date17 July 2009
CreatorsWang, Jen-feng
ContributorsJau-Wei Perng, Pei-Chung Chen, Chen-Wen Yen, Wei-Huang Lin, Yaw-Terng Su
PublisherNSYSU
Source SetsNSYSU Electronic Thesis and Dissertation Archive
LanguageCholon
Detected LanguageEnglish
Typetext
Formatapplication/pdf
Sourcehttp://etd.lib.nsysu.edu.tw/ETD-db/ETD-search/view_etd?URN=etd-0717109-161820
Rightsunrestricted, Copyright information available at source archive

Page generated in 0.0017 seconds