INCREASING RELIABILITY AND INFORMATION CONTENT

OF PULSE OXIMETRIC SAHS SCREENING ALGORITHMS

Nicole Gross, Jennifer Friedmann, Christophe Kunze, Wilhelm Stork

FZI Forschungszentrum Informatik, Haid-und-Neu-Strasse 10-14, Karlsruhe, Germany

Daniel Sánchez Morillo

, Antonio Leon Jimenez

, Luis Felipe Crespo Foix

Escuela Superior de Ingeniería Automática, University of Cádiz, Cádiz, Spain

Hospital Purta del Mar, University Hospital of Cádiz, Cádiz, Spain

Keywords: SAHS, Sleep apnea, Screening, Pulse oximetry.

Abstract: About 3% of people suffer from sleep apnea-hypopnea syndrome SAHS. SAHS is a sleep associated

respiratory disorder that negatively affects life quality and life expectancy. It is assumed that more than 80%

of SAHS concerned are neither diagnosed nor therapied. Reliable and easy-handling SAHS screening

systems are needed. Within this study, the reliability of pulse oximeter as a well-established, non-invasive

medical device is examined for SAHS screening. Reliability of existing SAHS screening algorithms will be

assessed. Hereby, the focus is on the influence of different desaturation detection strategies and the

dependence on thresholds. Critiques on pulse oximetry as SAHS screening device will be responded. In this

regard, guideline conform grey area integration in SAHS screening (concerning apnea-hypopnea index AHI

between 5 and 15) is recommended. In particular, as by grey area integration an improvement of convenient

SAHS screening algorithm reliabilities of about 7.3% in sensitivity and 8.7 % in specificity was achieved

even in the most reliable tested algorithm. In a final step, room for improvement of screening results

interpretation even without sleep medicine expert skills is indicated. In connection to this, possibilities of

short-term frequency analysis of SpO

data are demonstrated in its prospects for individualized SAHS

screening quality.

1 INTRODUCTION

One of the most common sleep disorders is the sleep

apnea-hypopnea syndrome SAHS. SAHS is a sleep

associated respiratory disorder, characterized by

repetitive breathing cessations. The hourly number

of respiratory breath arrests persisting more than 10

seconds (apnea) and reductions of respiratory flow

of at least 50% (hypopnea) is called the apnea-

hypopnea index AHI. AHI allows a classification of

apneic patients (cf. table 1). Threshold for

pathological AHI values is between 5/h and 15/h,

depending on the coexistence of documented

symptoms of excessive daytime sleepiness EDS.

(Downey, 2010) (S3-Guideline, Mayer 2009).

SAHS affects about 3% of people in the

industrialized countries (Young, 1993) whereat the

prevalence increases with male gender and age

Table 1: Guideline for apnea classification (Mayer, 2009).

AHI SAHS classification

AHI<5/h SAHS-healthy

5/h ≤AHI≤15/h

without EDS

SAHS-healthy

5/h ≤AHI≤15/h

with EDS

SAHS (mild)

15/h < AHI < 30/h SAHS (moderate)

AHI ≥ 30/h SAHS (severe)

(WHO World Health Organization). Standard

diagnosis for sleep apnea is the polysomnograpy, a

multi-sensorial physiological measurement to get a

complex sleep profil of the apnea suspect. As the

application of polysomnography is labor-intensive

and requires high technical expertise (Flemons,

2003), this examination is coupled with an overnight

stay in hospital or a aggreditied sleep laboratory.

Apart from high costs, required material and

upcoming waiting times for an examination, by this

438

Gross N., Friedmann J., Kunze C., Stork W., Sánchez Morillo D., Leon Jimenez A. and Crespo Foix L..

INCREASING RELIABILITY AND INFORMATION CONTENT OF PULSE OXIMETRIC SAHS SCREENING ALGORITHMS.

DOI: 10.5220/0003156104380445

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2011), pages 438-445

ISBN: 978-989-8425-35-5

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

way, patient’s inhibitions to undertake such an

examination must not be neglected.

Nowadays it is assumes that not even 10-20% of

moderate to severe SAHS patients are actually

diagnosed (Finkel, 2009), (Kapur, 1999), (Young,

1997), (Esteller, 2008), (Downey, 2010). Thereby an

early diagnosis and treatment of SAHS is proved to

reduce SAHS associated risk factors such as

fragmented structure of sleep, excessive daytime

sleepiness, loss of concentration and performance

and to augment the risk for cardio-vascular diseases

(Wantke, 2008). Untreated SAHS reduces life

quality and life expectancy for the patients

concerned. The requirement of reliable and easy-to-

use SAHS screening systems is obvious and already

persecuted on the market. Hence, this makes high

demands on the automatic analysis and screening

algorithms of pulse oximetric data.

Besides the medical requirements for reliability of

the screening results, there are several preconditions

that such screening systems should comply for its

application. As such, system should be applicable by

the patient himself after a short introduction by a

general practitioner. An automatic analysis and

interpretation of recorded screening data should be

possible as sleep medical interpretation skills of

recorded data cannot be presupposed (Netzer, 2001).

Widely-used and accepted for SAHS screening is the

overnight analysis of SaO

data by a non-invasive,

spectralphotometric pulse oximeter – wearable at the

fingertip or earlap. SaO

values determined by pulse

oximetry are called SpO

. Pulse oximetry is applied

in SAHS for the detection of desaturation in SpO

that are caused by apnea and hypopnea events. The

number of desaturations per hour is called the

desaturation index DI and correlates with AHI. The

intensity of respiratory desaturations is depending on

the apnea-hypopnea intensity and the inital SpO

value before the desaturation event.

Summarizing, pulse oximetry is already an easy-

to-use and cost-saving medical analysis device that,

based on automatic analysis of data, is sufficient for

reliable determination of apnea pretest probabilities.

(Wessendorf, 2002)(Mayer, 2009). But final

diagnosis, here the experts agree, is not to be

adviced by a screening system and should be

confirmed in a sleep laboratory.

2 METHODS

Within this study, existing pulse oximetric SAHS

screening algorithms based on time domain analysis

of SpO

[%] data are analyzed. Reviewing existing

algorithms, different strategies for apnea classifica-

tion are examined in their influence on reliabity and

comparability of screening results. Hereby, a main

focus of attention is given to the dependence on

thresholds and time windows under consideration.

Actual guidelines consider a grey area 5/h≤AHI≤15,

where an individual diagnosis is indicated. So, we

are going to weigh the pros and cons of grey area

integration into existing screening algorithms. As

future prospect, short-term-frequency analysis of

SpO

data is presented as a biosignal instrument to

sharpen the convenience and quality of patient

individual interpretation of SAHS screening results.

2.1 Pulse Oximetric Reference Data

The basis of the following comparisons is the

polysomnographic data of the University hospital

Puerta del Mar in Cádiz (Spain). It is the data of a

standard polysomnogram (Rechtschaffen, 1968)

(GES, 2005) including a fingertip pulse oximeter

(Erich-Jaeger, model number 70750A19, 2 Bytes

sample rate, range 0.0% - 100.0%). The sampling

rate of this pulse oximeter is 8 Hz. Desaturations are

defined to be at least 4 precentage points decreasing

in SpO

data without being limited on a certain time

interval. The patient collective includes 115 patients

of different age, gender and sleep apnea severity

code. DI values of patient collective are between 0

and 102; AHI is between 0.01 and 130. Figure 1

shows the correlation amongst the AHI and DI

values calculated by the hospital data.

Figure 1: Comparison of DI(hospital) and AHI(hospital).

The total correlation coefficient is 0.8436. In lower

AHI values, the correlation increases significantly.

In more specific terms, this means that the area of

AHI and DI <15 (classifier threshold for apnea

healthy and mild apnea patients) are correlated by a

coefficient of 0.9293.

For further calculations DI thresholds for apnea

classification are equalized to the AHI thresholds of

table 1 as there is a direct linear correlation between

DI and AHI values within the patient collective. This

assumption also coincides to preceding studies

INCREASING RELIABILITY AND INFORMATION CONTENT OF PULSE OXIMETRIC SAHS SCREENING

ALGORITHMS

439

(Torre-Bouscoulet, 2007) (Guideline of SEPAR,

2002). To quantify and assess the reliability of

different strategy and threshold compositions in

pulse oximetric SAHS screening algorithms, the

specificity (SP) and sensitivity (SE) (see equation 1)

of the combinations are compared as measurement

for the algorithms reliability.

SE =

right positive

right positive + false negative

SP =

right negative

right negative + false positive

Equation 1: Sensitivity (SE) and Specificity (SP).

2.2 Data Preprocessing

Before study analysis, pulse oximetric data

undergoes an artifact reduction that excludes

unphysiolgical SpO

values from data set.

Subsequently a moving average filter follows (cf.

equation 2). 2m+1 is the filter width, k is the index

of the actual data point and m equals 4 (corresponds

to 0.5 seconds of data record). The filter equation is:





∗

2 + 1









Equation 2: Moving average filter.

In relation to short-term analysis, SpO

data will

be downsampled to 1Hz and is filtered by a band

pass filter of bandwidth 0.015 Hz - 0.05 Hz. The

window length for short-term analysis is 60 seconds

with an overlap of 5/6. The estimation of power

spectral densities PSD in short-term frequency

analysis is made by the Yule-Walker algorithm.

Yule-Walker algorithm is an auto-regressive

function consisting of the last N measured values

y(t-1) to y(t-N) of the signal and a noise component

s(t) (cf. equation 3). s(t) is statistically normal

distributed. Mean value of noise is 0; standard

deviation is σ. The specified order of Yule-Walker

algorithm is 30.



(



)

=−





(

−

)

+()





Equation 3: Estimation of PSD via Yule-Walker

algorithm.

3 REALIZATION

By reviewing existing algorithms based on pulse

oximetric measurements, different strategies for

apnea classification can be identified. Within the

following study we will analyse the influence of

individual types of algorithms and thresholds on the

reliabity and comparability of screening results.

3.1 Reliability of Convenient SAHS

Screening Algorithms

Most existing and established algorithms refer to the

typical saw tooth desaturation course of SpO

data

as it results of precursory respiratory events. Figure

2 illustrates such a typical apnea desaturation event

including the subsequent resaturation process.

In principle, two accepted approaches in time

domain algorithms can be differed:

1. Desaturation based algorithms

2. Resaturation based algorithms

Figure 2: Typical course of an apnea caused desaturation.

Desaturation bases algorithms consider the slow

decrease of SpO

values, which occurs shortly after

a respiratory event. Frequently used thresholds are

3%, 4% or 5% for desaturation classifiers (Netzer,

2001). Often, there is also a time window indicated

to quantify the decreasing rate of SpO

data. Thus,

e.g. Rauscher (1991) examinded a time window of

40 seconds. Resaturation based algorithms focus on

the rapid reincrease of SpO

values at the end of

each respiratory event. To rebalance the oxygen

debt, breath frequency at the end of each SAHS

events is usually augmented. This results in an

accelerated increase of arterial oxygen saturation,

possibly with a short-term overshoot of SpO

above

of initial values (cf. figure 2). Approved thresholds

are increases of at least 3% within a time window of

10 seconds (Rauscher, 1991).

The following analysis is motivated by the

comparison of the contrast between such diverse

desaturation counting strategies. Desaturation (

↓) and

resaturation (

↑) appendages with different percentage

thresholds within predefined time periods will be

examined in relation to the screening reliability in

comparison to the AHI diagnosis. Reliability of

desaturation classifiers by at least 3% SpO

and 4%

SpO

decreases (desaturation method)

and

Intensity of

desaturation

[% SpO

]

Time of desaturation

rocess

Time of resaturation

rocess

Short-term

overshoot

of SpO

BIOSIGNALS 2011 - International Conference on Bio-inspired Systems and Signal Processing

440

accordingly at least 3% SpO

and 4% SpO

increases

(resaturation methods) are analyzed. Predefined time

periods are between 10 and 40 seconds generated by

corresponding moving time windows over the

recorded SpO

data. Double count of extended

desaturations in neighboring windows is avoided.

Precondition of a new event count is a previous

significant change of slope. In dependence on the

pathological defined AHI between 5 and 15 (cf.

table 1), the limit of a pathological amount of

desaturations per hour was center-defined by DI≥10

(q.v. GES, 2005) (cf. table 2).

Table 2: DI based SAHS classification.

DI classifier SAHS classification

DI < 10/h SAHS-healthy

DI ≥ 10/h SAHS (mild – severe)

The reliability of the DI classifier were assessed

in comparison to the hospital established SAHS-

classifier at AHI≥10 (cf. table 3). For this purpose,

SE and SP were determined. However, the resulting

reliability of all DI based SAHS classifier

modifications turned out to be not truly satisfying.

Even if SE tends to reliable percentages (maximum:

95.46%, mean: 84.89%), SP remains quite low

(maximum: 89.8%, mean: 67.49%) (cf. table 3). The

best composed result in respect to reliability was

reached by desaturation algorithm with desaturation

threshold at 4% in considered time windows of 30

seconds (SE=84.85%; SP=81.6%). According to

these analysis outcomes, we agree with the critics

complaining about the unsufficient specificity of

convenient pulse oximetric algorithms. Nonetheless

it may not be forgotten, that separating at DI=10 and

AHI=10 is a very hard threshold for SAHS

classification. Even actual guidelines (Mayer, 2009)

do not draw such a hard classifier line (cf. table 1).

Often, the algorithm results and the AHI diagnosis

from the hospital (AHI

hospital

) nearly agree and differ

in just a slight amount (total mean deviation of best

valued algorithm: -1.681). Nevertheless the final

classification sometimes fails because of the results

closeness to the classifier threshold.

3.2 Grey Area Integration

According to the German Guideline for apnea

classification, subjects with AHI between 5/h and

15/h are considered as a mild apnea patients if there

are coexisting EDS. If not, the subject is defined as

SAHS-healthy. Identification of EDS is depending

on a visit to a doctor and not possible by

conventional pulse oximetry screening methods.

Table 3: Reliability of desaturation and resaturation

counting thresholds.

SAHS classifier

algorithm

≥10

vs.

AHI

hospital

≥10

Mean deviation

Desaturation algorithms

Algorithm

SP (%) SE (%)

3%↓ 10s

77.6 81.82 1.910

3%↓ 20s

59.2 92.42 5.641

3%↓ 30s

57.1 93.94 5.458

3%↓ 40s

53.1 93.94 3.954

Algorithm

SP (%) SE (%)

4%↓ 10s

89.8 66.67 -5.264

4%↓ 20s

81.6 78.79 -1.785

4%↓ 30s

81.6 84.85 -1.681

4%↓ 40s

83.7 69.7 -4.133

Resaturation algorithms

Algortihm

SP (%) SE (%)

3%↑ 10s

59.2 89.4 4.476

3%↑ 20s

47.0 95.46 6.893

3%↑ 30s

47.0 95.46 6.015

3%↑ 40s

49.0 95.46 4.232

Algorithm

SP (%) SE (%)

4%↑ 10s

81.6 75.76 -2.316

4%↑ 20s

73.5 86.36 0.023

4%↑ 30s

69.4 86.36 -0.603

4%↑ 40s

69.4 87.88 -1.924

Applying this guideline considering the existence

of EDS, the previous algorithms will be reviewed.

DI values 5/h ≤DI≤15/h are reassigned to a new

defined grey area (cf. figure 3). The comparison to

hospital values keeps the limit at AHI=10/h.

It is obvious that both, SE and SP increase

significantly by grey are integration (cf. table 4).

The increase in SE values averages about 7%; and in

SP values 2.9%. This becomes apparent in figure 4.

Here, the ROC-curves of all tested algorithm

constellations are visualized in individual subplots.

Within these subplots, a comparison of algorithms

with and without grey area integration is provided.

The improvement of reliability is evident at prima

facie.

Considering the grey area integration, the best

reliability is calculated by the 4% decreasing

algorithm with a time window of 30 seconds. This is

equivalent to the non-grey-area algorithm analysis

result, but compared to the same algorithm without

the grey area classification group, SP improves

by8.7% to SP=90.32% and SE rises by 7.3% to

SE=92.16%. Both values, SP and SE above 90%

leads to the conclusion that pulse oximetry under the

described conditions might be a definitely valuable

and reliable device in screening of SAHS suspects.

INCREASING RELIABILITY AND INFORMATION CONTENT OF PULSE OXIMETRIC SAHS SCREENING

ALGORITHMS

441

Figure 3: SAHS screening classification algorithm including grey area.

Figure 4: ROC comparing algorithms with and without grey area integration.

Nonetheless, we would emphasize that subject

individual conditions are worth to be focused more

in screening systems. In this concern, a budding

prospect for pulse oximetric SAHS screening will be

presented within the following.

3.3 Prospects of Short-term Analysis in

SAHS Screening Interpretation

In the past, screening algorithms spent only little

attention to the intensity of desaturation events.

Thereby, desaturation events may vary between less

than 4% and up to 20% or more in SpO

. For

example, on the one hand, some hypopneas even do

not induce desaturations corresponding to

convenient definitions (e.g. decrease of at least 4%

in SpO

). On the other hand, and based on lower

initial SpO

values, the decrease of arterial oxygen

saturation proceeds quite fast even within short term

respiratory event. Hence, to determine the intensity

of SAHS caused desaturations, the duration of

desaturation events and the depth of SpO

decreases

has to be considered as an additional screening result

factor. In the fallowing we will derive a method

using short-term frequency analysis that permits to

depict the intensity of desaturation events in the

course of sleep in a quantitative, but easily

interpreted graphic image.

Examining the duration of single respiratory

caused desaturation events, it becomes apparent that

desaturations are predominantly of a duration

between 20 and 60 seconds (cf. figure 5).

BIOSIGNALS 2011 - International Conference on Bio-inspired Systems and Signal Processing

442

Figure 5: Duration of single desaturation events (e.g.

severe apnea patient).

Transfered to frequency domain, this results in a

minimum frequency of 0.05 Hz and a maximum

frequency of 0.0167 Hz for SAHS typical

desaturations (cf. equation 4).



_



_

20 

= 0.05 



_

_

60 

= 0.0167 

Equation 4: Frequency range of typical SAHS caused

desaturations.

Because of the saw tooth behaviour and thus

sinus likewise course of desaturations in SpO

, a

peak in power spectral density shapes between

0.0167 and 0.05 Hz (cf. (Gross, 2007) (Zamarrón

(peak between 0.014 Hz - 0.033 Hz), 1999)). The

peak height depends on depth and frequency of the

respiratory events within the considered time

window.

Figure 6: Intensity of desaturation events over sleep time.

The position on the frequency axis gives information

about the mean duration of the events within the

actual interval. Over the time axis, temporal

delimation of events is made possible.

Using short-term frequency analysis on recorded

SpO

data, the intensity of desaturation events can

be visualized in a clear and easy to interpret

individual SpO

sleep profil (cf. figure 6). Expressed

in simplified terms, it can be assumed, that the

higher the peaks, the more intensive are the

desaturation of the events are. And the more

frequent the peaks, the higher the desaturation

affected sleep time. As distinguished from time

domain presentation of SpO

data, non-SAHS

characteristic desaturations (e.g. long-term

desaturations caused by hypoventilation) are filtered

out of the desaturation intensity graphic image.

4 RESULTS

Within this study, we evaluated the reliability of

existing SAHS screening algorithms based on

overnight pulse oximetric records. We could provide

evidence that, in general, resaturation based

algorithms are the more specific algorithms and

desaturation based algorithms are the more sensitive

ones. Nonetheless, and according to actual

assessments on pulse oximetry in use for SAHS

screening, we have to summarize that the reliability

of convenient SAHS screening algorithms is

unsatisfactory. Such conclusion was reached turning

the attention to the maximum SE and SP results.

Thus, all screening algorithms constellations that

deliver a sensitivity of at least 85% do not reach a

specificity above 75%. None of the tested algorithms

stands out by both, a satisfactory high sensitivity and

specificity. Such unreliable result corresponds to the

comparable findings of many other critics on pulse

oximetric screening (Netzer, 2001), (Mayer, 2009) et

alii.

The mean difference of AHI and DI values was

generally quite small, especially in threshold near

SAHS patients, we assumed that the lack in

reliability must occur mainly because of the hard

SAHS classifier threshold. Classifier thresholds are

usually set at a fix and defined value like DI = 10.

Two patient groups result, those with DI < 10 being

the SAHS healthy classified and those with DI ≥ 10

being the SAHS suspects by pulse oximeter

screening. The hard threshold and the fact that AHI

and DI especially in moderate to severe SAHS

patients differ in a certain amount (cf. figure 1) leads

to unsatisfying SE and SP results. This assumption

AHI = 25.5

AHI = 89.9

Power Spectral Density Power Spectral Density

0.05Hz

0.0167Hz

INCREASING RELIABILITY AND INFORMATION CONTENT OF PULSE OXIMETRIC SAHS SCREENING

ALGORITHMS

443

was confirmed by including a grey area

characterized by the patient‘s individual need for

diagnosis. The integration of this grey area as an

additional classification group for apnea screening

could significantly improve the reliablity of SAHS

detection. The grey area was implemented in

reference to the actual German guideline that also

earmarks this area for requiring a patient‘s indivi-

dual classification in dependence on coexisting EDS

within these thresholds. The best result was obtained

by a decreasing type algorithm using a decrease

threshold of ≥4% within a moving time window of

30 seconds. The yielded reliability was SE=92.16

and SP=90.32 by a mean deviation of calculated DI

vs. real AHI of -1.68/h. In comparison to the same

algorithms without grey area consideration, this

showed an increase of 7.3% in SE and 8.7% in SP.

In a final manner, we animadvert that generally,

information about the intensity of desaturations gets

lost within existing SAHS screening algorithms.

Extended and deep desaturations (e.g. 40 seconds

and 15%) are not differed from shorter and light

desaturations (e.g. 20 seconds and 5%). In this

regard, we presented the prospects of short-term

frequency analysis of SpO

data. Thus, individual

sleep profiles of overnight desaturation

characteristics can be visualized in an easy to

interpret three-dimensional graph. This allows

drawing conclusions on intensity and repetitivity of

desaturation events, even by non sleep medicine

specialists.

5 DISCUSSION

According to sleep medicine guidelines, pulse

oximetry is a reliable medical device for SAHS pre-

test probability assessment; keeping in mind that is

not able to substitute a sleep medical diagnosis via

polysomnography (q.v. Netzer (2001), Wessendorf

(2002)). Thus, pulse oximetry is not able to detect

e.g. neither AHI nor EDS. Convenient SAHS

screening algorithms classify SAHS suspects

according to determinable DI values that can differ

from patients’ AHI. Possible causes of this diversity

of DI and AHI are multiple. For example there are:

- Hypopneas, that by definition are decreases

in breath flow, but that do not have to be

accompanied by pronounced desaturations

(Konietzko, 1998).

- Moving artifacts, that may lead to a over-

estimation of real DI values (Netzer, 2001).

- Physiological dependences like the initial

arterial oxygen saturation or the perfusion at

the point of measurement (Oczenski, 2008).

- Dependences on pulse oximeter model

(Zafar, 2005).

- Dependences of desaturation classifier

definition.

Nonetheless we demonstrated that the reliability

of SAHS pretest results yielded to trustworthy SE

and SP values ≥ 90% in comparison to AHI

diagnosis by polysomnography in sleep laboratory.

Against this background and with regard to the high

prevalence of SAHS we see a high, but still not

exploited potential of pulse oximetry in SAHS

screening and pretest application. Furthermore we

see the requirement and room for improvement of

the analyzability and interpretability of recorded

pulse oximetric screening data even by non-sleep

medicine experts. In this context, we presented

prospects of the short-term analysis of SpO

data for

improvement of SAHS screening by individual

SAHS sleep profiles considering desaturation

intensities and its temporal clusters.

REFERENCES

Downey, R., Gold, P. M., Wickramasinghe H., Medscape

2010, Obstructive Sleep Apnea

Esteller, E., Huerta, P., Segarra, F., Matiñó, E., Enrique,

A., Ademà, J. M., Acta Otorrinolaringol Esp. 2008,

Undiagnosed (corrected) cases of obstructive

sleepapnoea syndrome: a new reason for involvement

of otorhinolaryngologists, 59(2):62-9.

Finkel, K. J., Searleman, A. C., Tymkew, H., Tanaka, C.

Y., Saager, L., Safer-Zadeh, E., Bottros, M., Selvidge,

J. A., Jocobsohn, E., Pulley, D., Duntley, S., Becker,

C., Avidan, M. S., Sleep Med 2009, Prevalence of

undiagnosed obstructive sleep apnea among adult

surgical patients in an academic medical center,

10(7):753-8.

Flemons, W. W., Littner, M. R., Rowley, J. A., Gay, P.,

McDowell Anderson, W., Hudgel, D. W., McEvoy, R.

D., Loube, I. L., Chest 2003, Home Diagnosis of Sleep

Apnea: A Systematic Review of the Literature: An

Evidence Review Cosponsored by the American

Academy of Sleep Medicine, the American College of

Chest Physicians, and the American Thoracis Society,

124:1543-1579.

Gross, N., Kunze, C., Stork, W., Müller-Glaser, K., Rojas

Ojeda, J. L., Crespo Foix, L. F., Sánchez Morillo, D.,

León, A., BMT 2007, Schlafapnoeprädiagnose und -

klassifikation basierend auf nicht-linearer Analyse

pulsoximetrischer Daten.

Grupo español de sueño GES, Sociedad Española de

Neumología y Cirurgía Torácico SEPAR 2005,

Documento de consenso nacional sobre el síndrome

BIOSIGNALS 2011 - International Conference on Bio-inspired Systems and Signal Processing

444

de apneas-hipopneas del sueño (SAHS)

Kapur, V., Blough, D. K., Sandblom, R. E., Hert, R., de

Maine, J. B., Sullivan, S. D., Psaty, B. M., SLEEP

1999, The Medical Cost of Undiagnosed Sleep Apnea,

22(6):749-755.

Konietzko, N., Teschler, H., Freitag, L., 1998,

Schlafapnoe, Pathopysionlogie der obstruktiven

Schlafapnoe, 13.

Mayer, G., Fietze, I., Fischer, J., Penzel, T., Riemann, D.,

Rodenbeck, A., Sitter, H. Teschler, H., Somnologie

2009, S3-Leitlinie: Nicht erholsamer Schlaf /

Schlafstörungen.

Netzer, N., Eliasson, A. H., Netzer, C., Kristo, D. A.,

Chest 2001; Overnight Pulse Oximetry for Sleep-

Disordered Breathing in Adults: A Review. 120:625-

633.

Oczenski, W., Georg Thieme Verlag 2008, Atmen –

Atemhilfen – Atmphysiologie und Beatmungstechnik,

8.Auflage, p.370-371.

Rechtschaffen, A., Kales, A., 1968, Manual of

Standardized Technology, Techniques and Scoring

System for Sleep Association for the

Psychophysiological Study of Sleep

Rauscher, H., Popp, W., Zwick, H., Lung 1991,

Computerized Detection of Respiratory Events during

Sleep from Rapid Increase in Oxyhemoglobin

Saturation, 169:335-342.

Torre-Bouscoulet, L., Castorena-Maldonado, A., Baños-

Flores, R., Vázquez-García, J. C., Meza-Vargas, M. S.,

Pérez-Padilla, R., Arch Bronconeumol 2007,

Agreement between oxygen desaturation index and

apnea-hypopnea index in adults with suspected

obstructed sleep apnea at an altitude of 2240m,

43(12):640-54.

Wantke, F., J Kardiol 2008, Beatmungstherapie bei

Schlafapnoe – Indikation und Auswirkungen auf inter-

nistische Erkrankungen – ein Minireview, 15:364-8.

Wessendorf, T. E., Alymov, G., Wang, Y.-M., Stampa, J.

Thilmann, A. F., Teschler, H., Pneumologie 2002,

Pulsoximetrie als Schlafapnoescreening bei

Schlaganfallpatienten, 56:357-362.

Young, T., Palta, M., Dempsey, J., Skatrud, J., Weber, S.,

Badr, S., N Engl J Med 1993, The occurrence of sleep-

disordered breathing among middle-aged adults.

328(17):1230-5.

Young, T., Evans, L., Finn, L., Palte, M., Sleep 1997,

Estimation of the clinically diagnosed proportion of

sleep apnea syndrome in middle-aged men and

women, 20:705-706.

Zafar, S., Ayappa, I., Norman, R. G., Krieger, A. C.,

Wals-leben, J. A., Rapoport, D. M., Chest 2005,

Choice of Oximeter Affects Apnea-Hypopnea Index,

127:80-88.

Zamarrón, C., Romero, P. V., Rodriguez, J. R., Gude, F.,

Clinical Science 1999, Oximetry spectral analysis in

the diagnosis of obstructive sleep apnoea, 97:467-473.

INCREASING RELIABILITY AND INFORMATION CONTENT OF PULSE OXIMETRIC SAHS SCREENING

ALGORITHMS

445