Automatic scoring of apnea and hypopnea events using blood oxygen saturation signals. (September 2020)

• Record Type: Journal Article
• Authors: Rolon, R.E.; Gareis, I.E.; Larrateguy, L.D.; Di Persia, L.E.; Spies, R.D.; Rufiner, H.L.
• Is Part Of: Biomedical signal processing and control. Volume 62(2020)
• Publication Date: 2020-09
• Subjects: Pulse oximetry -- Apnea and hypopnea events -- Obstructive sleep apnea screening -- Structured dictionary learning -- Discriminant measures -- Multiclass classification problems; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2020.102062
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Only blood oxygen saturation signals are used to tackle the problem of individually differentiating between normal breathing, apnea and hypopnea events. A method for structured dictionary learning is used for dealing with the underlying problem. A multiclass discriminant measure, which is capable of highlighting important information of the dictionary atoms, is used. Significant improvements in obstructive sleep apnea screening are achieved by individually recognizing apnea and hypopnea events. The results encourage developing further intelligent technologies for portable screening devices, which can offer a new option for the proper detection of pathological patients. Abstract: The obstructive sleep apnea-hypopnea (OSAH) syndrome is a common and frequently undiagnosed sleep disorder. It is characterized by repeated events of partial (hypopnea) or total (apnea) obstruction of the upper airway while sleeping. To quantify the severity of the pathology, the Apnea Hypopnea Index (AHI) is used. This index is defined as the average number of apnea and hypopnea events per hour of sleep. Discriminating between these two types of events is a very challenging task and in fact most traditional methods fail to do it. A reliable recognition of such events would not only allow for an accurate estimation of the AHI index, but it would also provide useful information regarding the severity of the pathology, which is very important for clinical purposes. In this work we use a method for structured dictionary learning, which is found to be suitable for automatically differentiating between apnea and hypopnea using as a unique input blood oxygen saturation signals. The method is tested for both classification of segments and OSAH screening on the Sleep Heart Health Study database. For OSAH screening, a receiver operating characteristic curve analysis shows an average area under the curve of 0.934 and diagnostic sensitivity and specificity of 89.10% and 86.70%, respectively. These results represent important improvements with respect to all state-of-the-art procedures which where used for comparison purposes. They also provide a solid support for our conclusion that the method can be used for screening OSAH syndrome more reliably and conveniently, using only a pulse oximeter.