IRJET - Sleep Apnea Detection using Physiological Signals

INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056
VOLUME: 06 ISSUE: 06 | JUNE 2019 WWW.IRJET.NET P-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 1371
SLEEP APNEA DETECTION USING PHYSIOLOGICAL SIGNALS
Rekha S1, Dr. Shilpa R2
1M.tech Student, Dept of ECE, VVCE, MYSURU
2Associated Professor, Dept of ECE, VVCE, MYSURU
---------------------------------------------------------------------***-----------------------------------------------------------------
Abstract—Sleep apnea is one of the hypothetically serious
sleep disorder in which breathing frequently stops and
starts. The undiagnosed sleep apnea can be very severe and
hence rapid drops in blood oxygen level could occur during
which could increases possibility of advanced insulin
resistance and type 2 diabetes. However, several people are
unaware of their condition. The typical for diagnosing sleep
is an overnight polysomnography (PSG) in a committed sleep
laboratory. Since, these tests are costly and beds are limited
as educated staff needs to analyze the complete according.
An automatic detection method would permit a quicker
diagnosis and more patients to be analyzed. Hence detection
of sleep apnea is compulsory so that it could be treated. This
study established an algorithm that a short term event
extraction from Electrocardiography (ECG) signal and
combining neural network methodologies for automated
detection of sleep apnea. This study gives visual experiences
to the users by visual parameters such HRV measurements,
Poincare plot, Global and Local return map. This helps the
Doctor analyze whether the person is suffering from sleep
apnea or not.
Key Words: Sleep-apnea, Physiological signal, RR-
intervals, Heart-rate-variability measurement, neural
network.
1 INTRODUCTION
Sleep apnea is one of the most related sleep disorders
and is characterized by breathing pauses occurrence, also
identified as apneaic event, at night which leads to
recurrent awakenings [1]. It is naturally classified as either
obstructive sleep apnea (OSA) or central sleep apnea
(CSA).In obstructive sleep apnea is mutual kind of sleep
apnea which is affected by complete or partial obstructive
of higher airway and characterized by continue episodes of
shallow or paused breathing during sleep and decrease in
blood oxygen saturation. In central sleep apnea occurs
when the brain does not guide the signals which are needed
to breath. Sleep apnea is diagnosis using devices like
continuous positive air pressure (CPAP) machine,
Overnight Pulse Oximetry and Unattended Portable
Monitors. Untreated or undiagnosed sleep apnea can lead
to serious complication like danger of hypertension, cardiac
arrhythmia, heart attack and diabetes cancer and strokes
[2, 3]. Some of the studies are report that an estimated
49.7% of male and 23.4% of female adults suffer from sleep
disordered breathing, many cases remain undiagnosed as
patients are rarely aware of their condition [4].
A sleep apnea is diagnosed by overnight
polysomnography (PSG) recording at particular sleep
laboratory. This PSG records the cardiovascular
functioning, oxygen saturation, sleep status and multiple
physiological signals like ECG, PPG and EMG. Then using the
references of standard such as American Academy of sleep
medicine (AASM) [5] under the guidelines trained sleep
technical analyses the data of overnight and estimates each
fragment of the signals. Each fragment of the signal is then
noted as either OSA or CSA. Apnea-Hypopnea-Index (AHI)
represents the number of apnea and hypopnea measures
per hour and used to categorize patient into normal,
moderate or severe class. Due to limited quantity of beds
for PSG recording and the limited quantity of trained
technicians for examination waiting times will be extremely
long. The time taken to record ranges between 2 to 10
months in UK and in between 7 to 60 months in USA [6]. In
order to raise the total of the people that can be analyzed
and to reduce higher intra and inter-scorer variability
[7].The study need to assist the automated methods of
sleep technicians have stayed explored.
The wavelet transform is additional thought-provoking
method aimed for extracting waveform information in
which frequency resolution is good at lower frequency and
time resolution is high at higher frequency. In automatic
waveform detection is standard in wavelet based method
which is compared to a threshold for spike extraction.
In this work, a sleep apnea detection method is
proposed, based on a short term event extraction from ECG
signal and combining neural network methodologies for
automated detection of sleep apnea. By using the raw
physiological respiratory signals such as ECG signal, which
automatically learn and extract relevant feature and detects
the potential sleep apnea events.
2 RELATED WORK
2.1 Physiological signal
The sleep apnea diagnosis is done by using several
physiological signals such as ECG, EMG and PPG signals by
considering as input data. There are many devices to
diagnosis the sleep apnea like pulse oximetry,
polysomnography and Unattended Portable Monitors.
Analysis of those extents provides physicians with
important physiological information that may help them to
uncover fundamental dynamics of human health. It also
allows them to determine patient health state and choose
the right treatment. Respiration rate (RR) interval is stately

by the difference in the beat-to-beat intervals which is
identified as “variability of cycle length” or “variability of
RR”. Where R is a point resultant to the peak of QRS
complex of ECG wave and RR is the interval between
continuous. More definitely, the apnea period announces a
frequency element to the RR interval tachogram, which
resemble to the apnea duration. Heart rate variability
(HRV)has develop a appreciated non-invasive tool of
measuring the state of cardiovascular autonomic role in the
last decades and it has been frequently used in the analysis
of physiological signals in the various clinical and functional
situation .Over the recent years there was interest in heart
rate nursing without electrodes.
2.2 Sleep apnea detection
The sleep apnea effects were lead on human heart
constructed in electrical activity which is recorded [8].The
sleep is analyzed different types of sleep apnea were
detected such as obstructive sleep apnea and central sleep
apnea.
Several algorithms have been industrialized to detect sleep
apnea automatically by using one or two physiological
signal which is diagnosed by polysomnography device. In
adult’s sleep apnea were detected using overnight pulse
oximetry and unattended portable monitors.
However, various methods had been used in detecting sleep
apnea in which collective approach is used to interpret rule
based algorithms which gives information of certain epoch
signals and identified as having a sleep apnea or not [9].
Several approach have been commonly used such as
machine learning, kernel nearest neighbors (KNN), support
vector machine (SVM) and artificial neural network (ANN).
3 PROPOSED METHODOLOGY
The sleep apnea diagnosis is done using polysomnography
in sleep laboratory. Since the tests are expensive and
limited beds more number of patients cannot be analyzed.
These proposed methodologies can faster diagnosis and
also the more number of patients can analyze in which
automated detection of sleep apnea by short term event
extraction from physiological signals (ECG/PPG) and
combining back propagation in neural network method.
The ECG/PPG signals are loaded simultaneously. First the
person is tested by taking ECG signal as input data. If the
person does not have sleep apnea then testing is stopped.
If person is suffering from sleep apnea then testing done
using PPG signal as input data to validate.
The automated sleep apnea detection by physiological
signals such as ECG and PPG signals are considered in this
study. These physiological signals were trained using
neural network and tested using by the real time data.
Figure1. Illuminates to achieve the method for detection
of sleep apnea initially physiological signal are trained, in
which physiological data were collected from the physio-
net database. By using notch filter 50HZ noise frequency is
removed and wavelet transform is used to extract the
features from physiological signal and used back
propagation method in neural network which classifies the
Non -sleep disorder and sleep disorder. The testing is done
by using real time data in which gives the visualize output
by graphical user interphase.
Figure-1: Block diagram for sleep apnea detection.
Figure-2: Flow chart of sleep apnea detection.

Step 1: Data loaded
The real time data of the patient is collected. The obtained
data which is in analog form is converted into digital from.
Initially ECG data is consider as input data and loaded.
Step 2: Notch filter
The notch filter is used to remove the 50HZ noise
frequency in ECG signal with the sampling frequency of
1000HZ. The nyquist frequency is 500HZ, and notch filter
to nyquist ratio is 0.1.
Step 3: Wavelet
A wavelet transform technique analyzes the noise present
in ECG signal such as power line interferences and baseline
wander noises are removed by wavelet DB6.
Step 4: Breathing information
The breathing rate is measure of heart rate in which the
number of times the heart beats per minute.
Step 5: Global and local return map feature are
extraction
The HRV measurements are obtained and differentiated
for both local and global return map.
Step 6: Short time fragmentation
The ECG signal waveform is broken into desired range and
the functions of each smaller segment are identified.
Step 7: Local return map
This map foregrounds the occurrences of sleep apnea by
the nodes representation. If the nodes are closely spaced
then sleep apnea is not present. If the nodes are randomly
spaced then sleep apnea is present.
Step 8: HRV measurements and Poincare plot
The HRV measurements of various factors are obtained
and displayed both in global and local return map.
The Poincare plot analysis is made in the global return
map which identifies the RR interval. HRV increases when
the values are greatly spread while HRV decreases when
the values are lesser spread in global return map.
Step 9: Report
The obtained HRV measurement, visualize analyzes the
global and local return map output that are given to
Doctor. The Doctor checks the output and confirms
whether the person is suffering from sleep apnea or not.
Step 10: Validate with PPG
If the Doctor found that patient is suffering from sleep
apnea disorder using the ECG signal, once again test for
sleep apnea is conducted using PPG signal.
4 RESULTS AND DISCUSSION
In this study the automatic detection of sleep apnea by
short time event extraction from ECG/PPG signals which
are trained by the neural network using Matlab software.
The obtained results are represented by graphical user
interphase. If the person comes to known that is suffering
from sleep disorder (sleep apnea) then Doctor informs to
diagnose using PPG signal.
This study gives visual experience to the user by visual
parameters such as HRV measurements, Poincare plot,
global and local return maps.
The below simulation gives the output results of non-
sleep disorder, sleep disorder using ECG signal and also
PPG signal if the person is suffering from sleep disorder
(sleep apnea).
Table- 1:Non-sleep apnea HRV measurements of ECG
signal.
Non-sleep apnea HRV measures [ECG signal]
Global return
map
Local return
map
SDNN 0.3 0.2
PNN50 0.0 0.0
RMSSD 0.0 0.0
TRI 1.0 1.0
SD1|SD2 0.0|0.5 0.0|0.3
LF|HF 57.8|42.2 58.5|41.5
SD1/SD2 ratio 0.04 0.06
LF/HF ratio 1.37 1.41
TINN 16 16
Shift (-0.00,-0.00) (0.00,0.00)
IQR 0.77 0.61
RR HRV
median
0.90 0.75
Mean RR|HR 2|3.0325 2|2.5866

The table1 represents non-sleep apnea HRV
measurements by taking ECG signal as input data and
compares the global and local return map.
Table-2: Sleep apnea HRV measurements of ECG signal.
Sleep apnea HRV measures [ECG signal]
Global return
map
Local return
map
SDNN 24.2 29.3
PNN50 0.0 0.0
RMSSD 6.6 7.8
TRI 3.4 2.3
SD1|SD2 4.6|33.8 5.5|40.7
LF|HF 29.42| 70.58 NaN| NaN
SD1/SD2 ratio 0.14 0.13
LF/HF ratio 0.417 NaN
TINN 47 31
Shift (+0.22,+0.16) (+0.23,+0.12)
IQR 2.58 2.49
RR HRV
median
2.56 2.83
Mean RR|HR 48|1.244 48|1.246
The table 2 represents sleep apnea HRV measurements by
taking ECG signal as input data and compares the global
and local return map.
Table -3: Sleep apnea HRV measurements of PPG signal.
Sleep apnea HRV measures [PPG signal]
Global return
map
Local return
map
SDNN 0.5 0.3
PNN50 0.0 0.0
RMSSD 0.0 0.0
TRI 1.0 1.0
SD1|SD2 0.0|0.7 0.0|0.5
LF|HF 64.9|35.1 42.4|57.6
SD1/SD2 ratio 0.01 0.02
LF/HF ratio 1.85 0.74
TINN 16 16
Shift (+0.00,+0.00) (0.01,0.01)
IQR 0.71 1.00
RR HRV
median
0.81 1.13
Mean RR|HR 2|3.6173 1|5.0230
The table 3 represents sleep apnea HRV measurements by
taking PPG signal as input data and compares the global
and local return map.
(a)
(b)
(c)
Figure-3: Illustration input data of person. (a) ECG input
data for non-sleep apnea waveforms of breathing
represent zero fluctuations. (b) ECG input data for sleep
disorder represents fluctuations at certain point of sleep
duration. (c) PPG input data for sleep apnea represents
fluctuation at certain point of sleep duration.

(a)
(b)
(c)
Figure-4: Illustration the RR tachogram which further
encloses the mean RR and HR intervals, low and high
frequencies. (a) The RR tachogram for non-sleep apnea has
long RR and HR intervals and low and high frequencies
values are 57.83 and 42.2 percentages respectively. (b)
The RR tachogram for sleep apnea found that drastic
increased in RR interval and decreased in HR interval
values and obtained decreased low frequency i.e. 29.42
percentage and increased high frequency i.e. 70.58
percentage.(c) The RR tachogram which is further encloses
the mean RR and HR intervals, low and high frequencies
for sleep disorder by given PPG signal. RR tachogram for
sleep apnea using PPG signal as input data it is found that
RR interval is 2.00 and HR interval is 3.6173. Similarly
lower and higher frequencies values are 64.9 and
35.1percentage respectively.
(a)
(b)
(c)
Figure-5: Illustration the spectrum analysis of RR
tachogram. (a) Spectrum analysis of RR tachogram for non-
sleep apnea that implies low frequency to high frequency
ratios value is1.371. (b) Spectrum analysis of RR
tachogram for sleep apnea that implies low frequency to
high frequency ratios value is 0.412. (c) Spectrum analysis
of RR tachogram for sleep apnea by PPG signal as input
that implies low and high frequency ratio value is 1.85.
(a)
(b)

(c)
Figure-6: Illustration segmented HRV analysis. (a) ECG
signal segmented HRV analysis for non-sleep apnea in
which TINN is 16. (b) ECG signal segmented HRV analysis
for sleep apnea in which TINN is 47. (c) PPG signal
segmented HRV analysis for sleep apnea in which TINN is
16.
(a)
(b)
(c)
Figure-7: Displays local return maps. (a) Local return
produces the information regarding the each RR interval
predictions for non-sleep apnea in which the heart rate
variability is 0.75 and nodes are closer. (b) local return
produces the information regarding the each RR interval
predictions for sleep apnea in which heart rate variability
is 2.83 and nodes are farer.(c) Displays PPG signal local
maps provides the information regarding the each RR
interval predictors for sleep disorder nodes are also far
and heart rate variability is 1.13.
(a)
(b)
(c)
Figure-8: Displays global map Poin care plot. (a) Poincare
plot produces the information regarding entire RR
intervals for non-sleep apnea in which the heart rate
variability is 0.90.
(b)Poincare plot produces the information regarding
entire RR intervals for sleep apnea in which the heart rate
variability is 2.56. (c) Displays PPG signal imputed poin
care plot global map produces the information regarding
entire RR interval for sleep disorder in which the heart
rate variability is 0.81.
5 CONCULSION
The proposed work develops an algorithm that
demonstrates short time event extraction from ECG signal
and combining neural network methodology for automated
detection of sleep apnea. This study can analyze whether
the person is suffering from sleep apnea or not. If the
person is not suffering from sleep apnea then further

diagnosis is not required. If the person is suffering from
sleep apnea then further diagnosis is done using PPG signal
to validate. The proposed work gives subjective
information which helps Doctor to find out whether the
person is suffering from sleep apnea or not.
REFERENCES
[1] C. Guilleminault, A. Tilkian, and W. C. Dement, “The
sleep apnea syndromes,” Ann. rev. of medicine, vol. 27,
no. 1, pp. 465–484, 1976.
[2] V. K. somers, d. p. white, r. Amin, W. T. Abraham, F.
Costa, S. Daniels, J.S. Floras, C.E. Hunt, L. J. Olson, and
Others, “Sleep aplnea and cardiovasular disease,”
Circulation, vol.118, no, 10,pp . 1080-1111-2008.
[3] H. K. Yaggi, J. Concato, W. N. kernan, J. H. Lichtman, L.
M. Brass, and V. Mohsenin, "Obstructive sleep apnea as
arisk for stoke and death,” new Engalnd journal of
medicine, Vol. 27, no. 19, pp. 2034-2041, 2005.
[4] R. Heinzer, S. Vat, P. Marques-Vidal, H. Marti-Soler, D.
Andries, N. Tobback, V. Mooser, M. Preisig, A. Malhotra,
G. Waeber et al., “Prevalence of sleep-disordered
breathing in the general population: the hypnolaus
study,” The Lancet Respiratory Medicine, vol. 3, no. 4,
pp. 310–318, 2015.
[5] R. B. Berry, R. Brooks, C. E. Gamaldo, S. M. Harding, C. L.
Marcus, and B. V. Vaughn, “The AASM manual for the
scoring of sleep and associated events,” Rules,
Terminology and Technical Speciﬁcations, Darien,
Illinois, American Academy of Sleep Medicine, 2012.
[6] W. W. Flemons, N. J. Douglas, S. T. Kuna, D. O.
Rodenstein, and J. Wheatley, “Access to diagnosis and
treatment of patients with suspected sleep apnea,”
American journal of respiratory and critical care
medicine, vol. 169, no. 6, pp. 668–672, 2004.
[7] D. Bliwise, N. G. Bliwise, H. C. Kraemer, and W.
Dement, “Measurement error in visually scored
electrophysiological data: respiration during sleep,”
Journal of neuroscience methods, vol. 12, no. 1, pp. 49–
56, 1984.
[8] C. Guilleminault, R. Winkle, S. Connolly, K. Melvin, A.
Tilkian, Cyclical variation of the heart rate
in sleep apnoea syndrome: Mechanisms, and
usefulness of 24 h electrocardiography as a screening
technique, The Lancet 323 (8369) (1984) 126–131.
[9] G.Sannino,I.DeFalco,andG.DePietro,“Anautomaticrules
extraction approach to support osa events detection in
an mhealth system,” IEEE journal of biomedical and
health informatics, vol. 18, no. 5, pp. 1518– 1524,
2014.

IRJET - Sleep Apnea Detection using Physiological Signals

More Related Content

What's hot (18)

Similar to IRJET - Sleep Apnea Detection using Physiological Signals (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET - Sleep Apnea Detection using Physiological Signals