Content uploaded by Varun Pande
Author content
All content in this area was uploaded by Varun Pande
Content may be subject to copyright.
Content uploaded by Khaled Elleithy
Author content
All content in this area was uploaded by Khaled Elleithy
Content may be subject to copyright.
Content uploaded by Khaled Elleithy
Author content
All content in this area was uploaded by Khaled Elleithy
Content may be subject to copyright.
Development of OSA Event Detection Using
Threshold Based Automatic Classification
Laiali Almazaydeh, Khaled Elleithy, Varun Pande and Miad Faezipour
Department of Computer Science and Engineering
University of Bridgeport
Bridgeport, CT 06604, USA
{lalmazay, elleithy, vpande, mfaezipo}@bridgeport.edu
Abstract—Obstructive Sleep Apnea (OSA) is a very serious
sleeping disorder resulting in the temporary blockage of the
airflow airway that can be deadly if left untreated. OSA is not a
rare condition; in the US, from 18 to 50 million people, most of
them remain undiagnosed due to cost, cumbersome and resource
limitations of overnight polysomnography (PSG) at sleep labs.
Instead, automated, at-home devices that patients can simply use
while asleep seem to be very attractive and highly on-demand.
This paper presents a method for OSA screening and user
notification based on the respiratory recording and video
monitoring as a secondary system during sleep in order to alert of
the apnea event and help patient to recover.
Keywords: sleep apnea; PSG; respiratory signal; video
monitoring .
I. INTRODUCTION
A. Background and Motivation
Sleep is the circadian rhythm which is essential for human
life. Humans spend approximately one-third of their life asleep.
Sleep is necessary for optimal health; as we sleep, our body
repairs itself. During sleep, blood pressure fluctuates, heart rate
slows down, hormone fluctuations occur, muscles and other
tissues relax and repair, and the replacement of aging or dead
cells occur. Without sleeping, we simply do not function as
well as we can [1].
The 25 year old field of sleep medicine, has now covered
84 kinds of sleep disorders, including the most common ones
such as narcolepsy, insomnia, sleep apnea, and restless leg
syndrome [2].
Sleep apnea (SA) is becoming a most common respiratory
disorder during sleep, which is characterized by cessations of
airflow to the lungs. These cessations in breathing must last
more than 10 seconds to be considered an apnea event. Apnea
events may occur 5 to 30 times an hour and may occur up to
four hundred times per night in those with severe SA [3].
There are two major types of SA. One type of SA is known
as obstructive, which is generally caused by a blockage of the
airflow airway. Central sleep apnea (CSA) is the other type of
apnea, which occurs when the brain’s drive to breathe is
reduced. Most cases of CSA are mixed, meaning that it is often
along with OSA. However, OSA is more common in the
general population than CSA [3].
In fact, SA is associated with a major risk factor of health
implications and increased cardiovascular diseases and sudden
death. It has been linked to irritability, depression, sexual
dysfunction, high blood pressure (hypertension), learning and
memory difficulties, stroke and also heart attack [4].
The most frequent night symptoms of SA can include
snoring, nocturnal arousals, sweating and restless sleep.
Moreover, like all sleeping disorders, symptoms of sleep apnea
do not occur just during the night. Daytime symptoms can
range from morning headaches, depression, impaired
concentration and excessive sleepiness which cause mortality
from traffic and industrial accidents. However, these symptoms
are not definitive to detect SA syndrome. Nowadays, PSG is a
standard testing procedure to diagnose OSA. Complete PSG
includes the monitoring of the breath airflow, respiratory
movement, and oxygen saturation (SpO
2), body position,
electroencephalography (EEG), electromyography (EMG),
electrooculography (EOG), and electrocardiography (ECG).
Therefore, a final diagnosis decision is obtained by means of
medical examination of these recordings. Nevertheless, the
PSG process is a complex, expensive and time consuming
procedure due to the need of many physiologic variables using
multiple sensors attached to the patients [5].
To summarize, new simplified diagnostic methods and
continuous screening of OSA is needed in order to have a
major benefit of the treatment on OSA outcomes. In this
regard, we present a work based on breathing detection and
video monitoring that will be used in a larger real time system
for OSA diagnosis. The objective of the system is to alert a
patient who might be subject to an apnea attack.
B. Paper Organization
The rest of this paper is organized as follows. In Section II,
we glance at a variety of OSA detection methods. Section III
contains an overview of our system, including the detection
algorithm steps, and details on the analysis methodology of the
paper. In Section IV, we provide details of the experiment and
the results of our system. Finally, Section V concludes this
paper regarding the potential usefulness of our system, and
highlights directions for future research.
II. R
ELATED WORK
Over the past few years, most of the related research has
focused on detecting OSA through statistical features of
different signals such as thorax and abdomen effort signals,
nasal air flow, oxygen saturation, electrical activity of the heart
(ECG), and electrical activity of the brain (EEG).
In our previous published research, we developed a model
based on a linear kernel Support Vector Machines (SVMs)
using a selective set of RR-interval features of the ECG signal
[5]. In addition, in [6] we further developed a Neural Network
(NN) as a predictive tool for OSA using oxygen saturation
signal (SpO
2) measurements obtained from pulse oximetry. The
authors in [7] assessed a compendium of features extracted
from EEG and Heart Rate Variability (HRV) data using
advanced signal processing approaches. The detection
algorithms in [6-8] assessed and validated their results based on
polysomnographic data that was available online from the
physionet database, which offers free access to Apnea-ECG
Database [8]. Results show that trends detected by those signals
features could distinguish and annotate apnea events, and those
methods could prove helpful in computer aided detection of
sleep apnea.
Nowadays, much of the current apnea research is being
done on providing portable devices that monitor those
experiencing apnea during the day by alerting them when they
stop breathing. The device could act as an inexpensive and
convenient way for doctors to diagnose SA patients and as a
means for collecting data on apnea sufferers to determine the
severity of the condition once diagnosed. More specifically,
this may help in the initial assessment of patients with
suspected OSA in order to prioritize patients. Patients with
utmost need of treatment will go through complete PSG
recordings within a sensible time frame; meanwhile those who
are free of apnea symptoms will avoid the cumbersome
procedure [9].
Various portable monitor devices already exist in the
market. ApneaLink
TM
Plus Home Sleep Apnea Test Device is
one of the portable home sleep test diagnostic devices that
records up to four channels of information: respiratory effort,
pulse, oxygen saturation and nasal flow. The patient can sleep
normally while ApneaLink
TM
Plus monitors his/her sleep,
checking breathing patterns and the amount of oxygen in
his/her blood and recording possible apneas or other breathing
abnormalities [10]. Also, WM ARES is a home sleep test
device that records heart rate, airflow, respiratory effort and
oxygen saturation [11].
When the patient wakes in the morning, after removing the
tube from the nose and the tape and sensor from the finger,
he/she returns the device to the clinician for analysis. The
device contains a detailed record of the patient’s personal sleep
patterns, which can be downloaded, analyzed and processed in
the clinician’s computer. The clinician will then identify if the
person is suffering from sleep apnea.
In the current study we have designed a home monitoring
system for OSA detection and user notification through the use
of a small microphone placed on the neck of the patient to
differentiate normal and disrupted breathing, and movement
monitoring that will serve as a secondary means of detecting
apnea events.
Based on the mentioned related works in this area, our
contribution in this work is developing a real time system that
is used to alert the patient during an apnea event and remind
him/her to breath.
III. N
EW APPROACH
In this work, the goal is to capture and precisely identify
breathing sounds and alert the user/patient whenever the
acoustic signal of respiration is paused.
It is important to note that processing and analyzing breath
generally includes several fields of analysis. In what follows,
our detection system design is described.
A. Overview
The overall design involves acquiring sound from a
microphone. This signal is then processed to detect
abnormalities in breathing or breathing cessations. Moreover,
as a secondary system, image data is continually collected from
the environment through webcam. When an apnea event
occurs, that is, if the patient is without breath for longer than 15
seconds, and the conditions of surrounding environment is
changing, an alarm will be raised immediately.
B. Apnea Detection Prototype
Figure 1 provides an overview for the apnea monitoring
system.
The breathing sounds are recorded using a small
microphone. We have integrated the microphone with a large
stethoscope-like diaphragm in place with a soft band that is
fastened gently around the patient’s neck to ensure that the
sound recordings take place comfortably during the patient’s
sleep. Figure 2 shows a soft neck band against user view.
During the initialization of the detection, three main
modules are initialized simultaneously; the peak programme
meter (PPM), the analyzer and the detector. The initial sound
data is recorded through a signal recorder, which is a peak
programme meter (PPM). PPM in general, has the ability to
measure peak levels for any input sound. This allows us to have
the first raw information about the patient’s breath cycles and
the peak threshold values for an apnea attack. The values from
the PPM meter are named as wav 1 (audio data 1). The wav 1
values are sent to a filter to remove the noise and is then stored
in the data-base. While a copy of the sound is stored in the
database, another copy of the sound file is sent to the detector;
this is a secondary data for detection (system fail trigger) of
sleep apnea. The initial data of the PPM are then sent to the
analyzer for the development of the classifier. Here, an expert
sound analyst or a trained doctor can use the initial input to
build the classifier based on the first sound file, namely wav 1.
The development of the classifier is the second simultaneous
step that is running once our detection process has begun. In
the second step, the breathing sounds from the microphone are
passed through the analyzer which shows the graphical user
interface of the sound wave. These sound waves can then be
used to adjust the patient’s breath cycle, thus making it possible
to design a breathing classifier. This means we can set the
pattern for regular breathing and a pattern for irregular
breathing.
Figure 1. The framework for OSA detection.
Figure 2. The microphone in the stethoscope and its neck band.
Both these patterns are sent through a filter to remove noise
and the final data is named wav 2, which is the second sound
file for our detection process. This sound is stored in the data
base as classified data. This data is more specific for the patient
and is improved throughout as more and more data is initialized
for a specific patient. This way we have a case history of
sound files for a patient and a specialized classifier for each
individual patient’s apnea disease. The final step in our
detection process and the third simultaneous process is the
detector module. This detector module receives values from
either the PPM which is a secondary value in case a classifier
for a patient is not set (our system fail trigger) or receives the
primary value from the analyzer module for the detection.
During the final step of the detection process, if the threshold
values input based on the classifier are exceeded, then an alarm
is raised. In case the patient does not have the microphone
module around his neck and the system is in the monitoring
mode, then it might raise an alarm. To avoid this, there is a
video monitoring tool that judges the patient’s presence.
However, this module is not a primary requirement to our
detection process. In case the patient does not have an apnea
attack for a time period of thirty seconds, then the whole
detection process will run again with the same classified data
values. This is a never ending process that keeps on monitoring
the patient throughout his sleep cycle while collecting useful
data for further analysis and better classification for the future.
IV.
EXPERIMENT AND RESULTS
The complete apnea detection system is implemented in
Visual C# in MS-Windows platform. The Olympus ME-52W
noise canceling microphone to capture breath and a standard
web-camera with 60/sec to capture video frames are used. The
experimental results were conducted on a Pentium 2.4 GHz
computer system.
Figure 3 shows the interface used in our sound analyzer
module, where the user can study and classify the breathing
cycle of the patient.
Figures 4 and 5 respectively show the results of SA
detection from the abnormal breathing and movement
detection. The graphical results screen illustrates the results in
real time based on the values fed to our application from
module 1, module 2 and module 3. The resultant graph is based
on the RMS (root mean square) of the first three graphs. A user
can access this application to get the analytical report as a print
out with the inbuilt function.
The developed system was confirmed to allow for a
comfortable, continuous and real time monitoring of sleep
apnea disorder at home, which can trigger an alarm if the apnea
is detected.
Figure 3. The Analyzer
Figure 4. The detector
Figure 5. The SA detection framework results graph.
IV. CONCLUSIONS AND FUTURE WORKS
In this work, we presented a sleep apnea detection system
which is capable of performing real time monitoring and
visualization of sleep data. In the future, we plan to add other
physiological data like SpO
2, and then send sleep data
wirelessly via Bluetooth to a nearby cell phone for processing
and storage. We plan to build a cloud application to keep
health information with personal information of the
individual’s records, which will offer access to a large pool of
sleep data for further investigations.
Our aim is to establish a simple and helpful at home OSA
screening system by which data can be uploaded to cloud and
transmitted to a hospital for sleep data analysis.
This research direction of accumulating patients’ data will
augment the efforts in this challenging field through providing
benchmark data that can be used by researchers to enhance
their used mechanisms and tools.
R
EFERENCES
[1] W. Mendelson, Human Sleep: Research and Clinical Care. Plenum
Medical Book Company, New York and London.
[2] Sleep Disorder Overview. www.neurologychannel.com
.
[3] Sleep Disorders Guide. www.sleepdisorderguide.com
.
[4] G. Angius and L. Raffo, “A Sleep apnoea keeper in a wearable device
for continuous detection and screening during daily life,” Computers in
Cardiology, vol. 35, pp. 433-436, Sep. 2008.
[5] L. Almazaydeh, K. Elleithy, and M. Faezipour, “Detection of obstructive
sleep apnea through ECG signal features,” in Proceedings of the 2012
IEEE International Conference on Electro Information Technology
(IEEE eit2012), pp. 1-6, May. 2012.
[6] L. Almazaydeh, M. Faezipour, and K. Elleithy,” A Neural network
system for detection of obstructive sleep apnea through SpO
2 signal
features,” (IJACSA) International Journal of Advanced Computer
Science and Applications, vol. 3, no. 5, pp. 7-11, Jun. 2012.
[7] E. Estrada and H. Nazeran,”EEG and HRV signal features for automatic
sleep staging and apnea detection,” in Proceedings of the 20
th
IEEE
International Conference on Electronics, Communications and
Computer (CONIELECOMP), pp. 142-147, Feb. 2010.
[8] PhysioNet, www.physionet.org
.
[9] T. Shochat, N. Hadas, M. Kerkhofs, et al., “The SleepStrip
TM
: an apnoea
screener for the early detection of sleep apnoea syndrome,” European
Respiratory Journal, vol. 19, pp. 121-126, 2002.
[10] www.resmed.com
.
[11] M. Stuart,”Sleep apnea devices: the changing of the guard,” START-UP
journal, vol. 15, no. 10, pp. 1-8, Oct. 2010.
