Automatic Wheeze and Respiratory Phase Detectors to Evaluate

Respiratory Physiotherapy in LRTI

A Preliminary Study

João Dinis

1,2

, Ana Oliveira

, Cátia Pinho

1,2

, Guilherme Campos

, João Rodrigues

and Alda Marques

School of Health Sciences, University of Aveiro (ESSUA), Campus Universitário de Santiago, Aveiro, Portugal

Institute of Electronics and Telematics Engineering of Aveiro (IEETA), University of Aveiro,

Campus Universitário de Santiago, Aveiro, Portugal

Keywords: Wheezing, Respiratory Phases, Respiratory Physiotherapy, Lower Respiratory Tract Infection.

Abstract: Respiratory physiotherapy is a gold standard intervention for chronic respiratory conditions. However, its

application in acute respiratory diseases (e.g., LRTI) is not well established. The objective and reliable

measurement of adventitious lung sounds (ALS), such as wheezes, has the potential to contribute to

respiratory physiotherapy evidence base. This paper reports on the implementation of reliable and published

automatic wheeze and respiratory phase detectors to assess wheezing parameters pre/post respiratory

physiotherapy treatment in patients with LRTI. Twenty patients with LRTI were randomly allocated to

control group, which received standard medication treatment, or experimental group, which received

standard medication plus respiratory physiotherapy treatment. Respiratory sounds were recorded in seven

chest locations. Wheeze parameters, namely occupation rate, main frequency, duration and type were

obtained per respiratory phase. Wheeze occupation rate was statistically significantly reduced in both

groups following treatment (p<0.001). There was a greater reduction in wheeze occupation rate in the

experimental group reaching statistical significance for the inspiratory phase (p=0.019). This promising

result indicates the potential value of respiratory physiotherapy in LRTI. It also highlights the potential to

use acoustic methods to establish respiratory physiotherapy efficacy.

1 INTRODUCTION

Lower respiratory tract infection (LRTI) covers a

wide range of diseases from a mild mucosal

colonisation or infection, an acute exacerbation of

chronic bronchitis/chronic obstructive pulmonary

disease (COPD), to an overwhelming parenchymal

infection such as community acquired pneumonia

(CAP) (Woodhead et al., 2011).

It is estimated that the annual incidence of adult

people with LRTIs consulting healthcare providers

ranges from 8-124 per 1000 population in Europe

(Ward and Ayres, 2000) and more than 5 million

cases of CAP occur annually in the United States of

America, especially in the winter months (Graham,

2008). Any age group can be affected however,

LRTI is more common in those under 5 and above

45 years old (Graham, 2008).

Respiratory physiotherapy has been recognised

as an important component in the treatment of

respiratory patients. Evidence of benefit has been

demonstrated in chronic respiratory conditions

(Garrod and Lasserson, 2007). However, there is a

need to establish efficacy in acute respiratory

diseases (e.g., LRTI).

It is widely accepted that adventitious lung

sounds (ALS), namely crackles and wheezes,

contain important information about pulmonary

dysfunctions (Laennec, 1935). Wheezes have been

the most common type of ALS investigated for

diagnostic purposes using the stethoscope (Earis and

Cheetham, 2000).

Wheezes are pitch-based sounds sustained for

longer than 100 ms with frequencies above 100 Hz.

It can be classified as monophonic (single

frequency) or polyphonic (multiple frequencies) and

occur mostly during expiration, however they can

also be heard during inspiration in more severe cases

(Sovijärvi et al., 2000). This ALS can be heard in

several diseases involving narrowing of airway

calibre (Meslier et al., 1995). Although COPD and

asthma are the main respiratory diseases presenting

233

Dinis J., Oliveira A., Pinho C., Campos G., Rodrigues J. and Marques A..

Automatic Wheeze and Respiratory Phase Detectors to Evaluate Respiratory Physiotherapy in LRTI - A Preliminary Study.

DOI: 10.5220/0004246702330238

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2013), pages 233-238

ISBN: 978-989-8565-37-2

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

wheezes (Waris et al., 1998), this type of ALS also

contributes to the diagnosis and monitoring of LRTI

(Woodhead et al., 2011).

According to the European Respiratory Society

(ERS) guidelines (Charbonneau et al., 2000), the

percentage of the respiratory cycle occupied by

wheezes is of special interest - the higher the

percentage the more severe the disease is (Sovijärvi

et al., 2000). Previous studies have shown an

association between the degree of bronchial

obstruction and the proportion of the respiratory

cycle occupied by wheezing (Baughman and

Loudon, 1984). However, additional measurements

such as the number of wheezing peaks, their main

frequencies, duration, timing in respiratory cycle and

location of the recording (chest wall or trachea) can

also be relevant and should be calculated if possible

(Piirilä et al., 2000). In order to address these

parameters an automatic acoustic approach is

desirable.

Several algorithms have been proposed to detect

wheeze parameters (Taplidou and Hadjileontiadis,

2007); (Qiu et al., 2005) and respiratory phases (Huq

and Moussavi, 2010); (Yildirim et al., 2008).

Taplidou and Hadjileontiadis’ (2007) algorithm has

been reported as the one with the best performance

(Oliveira et al., 2011). For respiratory phase

detection, Huq and Moussavi’s (2010) algorithm is

the most recent and overcomes limitations reported

by previous studies.

Therefore, this paper reports on the

implementation of a reliable and published

automatic wheeze and respiratory phase detectors to

assess wheezing parameters pre/post respiratory

physiotherapy treatment in patients with LRTI.

2 METHODS

A randomised controlled trial was conducted.

Ethical approval was obtained from the Ethics

Committee of Hospital Infante D. Pedro, Aveiro,

Portugal.

2.1 Procedures

Patients were eligible for the study if they presented

with cough and at least one of the following

symptoms: sputum, dyspnoea, wheezes or chest pain

(Woodhead et al., 2011), at the emergency

department of the Hospital Infante D. Pedro (Aveiro,

Portugal). Twenty participants (10 males) diagnosed

with LTRI by the physician, according to the LRTI

guidelines (Moher et al., 2010); (Woodhead et al.,

2011), were recruited for the study. A simple

randomisation (Moher et al., 2010) was use to

allocate patients to the control group or experimental

group.

The control group was treated with standard

medication, i.e. antibiotics, and the experimental

group received the same standard medication plus

respiratory physiotherapy for acute respiratory

conditions. A physiotherapy protocol was carried out

three times per week (British Thoracic Society,

2001) for 3 weeks (Woodhead et al., 2011)

accomplishing a total of 9 sessions. Each session

lasted on average 90±15 minutes (American College

of Sports Medicine, 2006). The intervention protocol

consisted of: i) breathing retraining techniques to

reduce energy costs of breathing and dyspnoea

(American College of Sports Medicine, 2006); ii)

inspiratory techniques such as incentive spirometry

to increase pulmonary expansion (Weiner et al.,

1997), prevent atelectasis and aid at sputum

clearance (Postiaux, 2004); iii) airway clearance

techniques such as the active cycle of breathing to

mobilize and clear excess bronchial secretions

(Pryor and Prasad, 2008); iv) exercises for thoracic

mobility, expansion and flexibility to increase

pulmonary volumes; v) aerobic training (walking

and cycling) at 60-80% of the patient maximal

cardiac frequency to increase tolerance to physical

activity and improve the physical fitness of the

patient (American Association of Cardiovascular

and Pulmonary Rehabilitation, 2006), vi)

educational support about the disease and lifestyles

to ensure on going effective intervention and to

provide the patient with some control over the

disease and vii) prescription of home exercises.

All treatment sessions were held in a well-

equipped room at University of Aveiro.

2.2 Data Collection

Data were collected by two researchers in a clinical

setting within 24 hours of hospital presentation and

after 3 weeks of treatment, the time taken to recover

from a LRTI (Woodhead et al., 2011).

2.2.1 Demographic, Anthropometric and

Lung Function

Demographic and anthropometric data was first

collected (height and weight to calculate the body

mass index). Lung function evaluation involved the

collection of forced expiratory volume in 1 second

(FEV

) and forced vital capacity (FVC) with the

spirometer MicroLab Micro Medical 36-ML3500-

HEALTHINF2013-InternationalConferenceonHealthInformatics

234

MK8, UK, following the ERS guidelines (Miller,

2005).

2.2.2 Respiratory Sounds

Respiratory sound recordings were performed

according to the Computerized Respiratory Sound

Analysis (CORSA) guidelines for short-term

acquisitions, in a clinical room, i.e., participants

were in a seated-upright position and lung sound

data was collected using seven modified analogue

stethoscopes (Classic II S.E., 3M™ Litman®, St.

Paul, MN, USA). Each stethoscope was attached to

the body using an adhesive tape (Leukosilk®, BSN

Medical GmbH, Hamburg, Germany) in seven chest

locations, i.e., trachea, left and right anterior, lateral

and posterior regions. Respiratory sounds were

collected by custom-made microphone and

preamplifier circuit (Intelligent Sensing

Anywhere®, Coimbra, Portugal) inserted into the

main tube of each stethoscope. The resulting

analogue signals were further amplified and

converted to digital by a multi-channel audio

interface (M-Audio® ProFire 2626, Irwindale, CA,

USA). The signal was converted with a 24-bit

resolution at a sampling rate of 44100 samples per

second in each channel and recorded in wave format

on a laptop computer. A diagram of the recording

setup is shown in figure 1.

Figure 1: Diagram of the recording setup for one

stethoscope.

The average time between the first (pre-

treatment) and the second respiratory sound

recording session (post treatment) was 22.0±8.8

days for the control group and 22.8±3.1 for the

experimental group. Three repetitions per participant

(20 seconds each) were performed in each time.

2.3 Automatic Detection Algorithms

Taplidou and Hadjileontiadis (2007) automatic

wheeze detector and Huq and Moussavi (2010)

automatic respiratory phase detector were

implemented, as they have been shown to be reliable

(overall performance of 94.6% (2007) with an

accuracy of 93.1% (2010)). The combination

between these two algorithms allowed the

calculation of the wheeze occupation rate in each

respiratory phase (i.e., inspiration, expiration) of the

recorded signals.

The following sections present a brief description

of these algorithms.

2.3.1 Wheeze Detection Algorithm

Wheezes were detected using the algorithm

described by Taplidou and Hadjileontiadis (2007).

This algorithm is based on the Short-time Fourier

transform (STFT), proposed by Gabor (Gabor,

1946), which is a classical method for analysing

non-stationary signals. This technique is a Fourier-

related transform used to determine the sinusoidal

frequency and phase content of local sections of a

signal as it changes over time. The Fourier transform

of the resulting signal is taken as the window is slid

along the time axis, resulting in a two-dimensional

representation of the signal, where x





 is the signal

and ω denotes the spectral window.

STFT









≡X



τ,ω



 x





.ω



tτ









In the implemented algorithm, the signal is digitally

filtered (band pass 60-2100Hz, order-8 Butterworth)

and resampled (to 5512s

-1

) before the STFT

calculation. To remove noise from the STFT signal,

a smoothing procedure based on box filtering, also

known as mean-filtering, estimates the trend of the

frequency content of the windowed signal at each

time instant. Peaks higher than a specific magnitude

threshold are then selected. These peaks are then

classified as wheezes or non-wheezes according to a

set of criteria that include local maxima, peak

coexistence and continuity in time.

The algorithm allowed the calculation of

different parameters, e.g., starting and ending time

as well as fundamental frequency of each detected

wheeze. It was also possible to classify the wheeze

according to its type (monophonic or polyphonic).

2.3.2 Respiratory Phase Detection

Algorithm

For the respiratory phase detection, an algorithm

using only tracheal sounds was implemented (Huq

and Moussavi, 2010). Because of the synchronized

multi-channel acquisition, the detected phases and

onsets on the tracheal sounds were used to calculate

the wheeze occupation rate in the other six places,

where the acoustic signal was acquired.

AutomaticWheezeandRespiratoryPhaseDetectorstoEvaluateRespiratoryPhysiotherapyinLRTI-APreliminaryStudy

235

Similarly to the wheeze detector algorithm, the

signal was firstly digitally filtered (band pass 150-

800Hz, order-10 Butterworth filter) and resampled

(to 10240 s

-1

). The selected filtering band was used

to minimise the effect of heart sounds and high

frequency noises. In this algorithm, several

parameters were collected from the duration, volume

and shape of the tracheal breath sound envelope in

each phase. For this purpose the logarithmic

variance (LV) of the filtered sound signals was

calculated. As the LV of the breath sounds resemble

a fully rectified flow signal, respiratory onsets (i.e.,

starting sample of a respiratory phase) can also be

detected. Using the majority-vote of parameters

between adjacent phases, they can be classified as

inspiration or expiration.

2.4 Statistical Analyses

Statistical analysis was conducted using SPSS®

19.0. Differences between parameters in the first

(pre-treatment) and the second respiratory sound

recording (post treatment) were explored with paired

samples T-test. Wheeze occupation rate (R), main

frequency (F) and type of wheeze (T) evaluated on

both inspiratory and expiratory phase recordings

were analysed. R value was established in the range

zero to one (where 0 was given when no wheezes

were detected and 1 when the respiratory phase was

fully occupied). The T was classified as 0 if

monophonic or as 1 if polyphonic.

Statistically significant differences between

groups (control vs. experimental group) at each

parameter assessed on pre and post treatment was

explored to evaluate the impact of the respiratory

physiotherapy. For this purpose also an independent

samples T-test analysis was performed.

Data were expressed as mean and standard

deviation (Mean±SD). Significance level was set at

p<0.05.

3 RESULTS

A total of twenty participants (10 males) diagnosed

with LTRI enrolled in this pilot study. Eleven

patients (4 males) composed the control group while

the experimental group was composed by 9 patients

(6 males). The sample is characterised in Table 1.

Paired sample t-test results for pre-post treatment

analysis on control and experimental groups are

present in Table 2. A statistically significant

decrease was observed in both inspiration and

expiration wheeze occupation rate in both groups.

Table 1: Sample characterisation.

Age (yrs) BMI (kg/m

) FVC

(%) FEV

1-PP

(%)

CG 52.9 ± 18.3 26.1 ± 5.2 75.7 ± 21.6 72.2 ± 29.8

EG 49.9 ± 23.2 23.4 ± 4.6 62.6 ± 25.9 62.2 ± 29.0

T 56.0 ± 13.7 24.9 ± 4.9 69.8 ± 23.9 67.7 ± 28.2

CG – Control group; EG – Experimental group; T- Total; BMI –

Body mass index; FVC

– percentage predicted of forced vital

capacity; FEV

1PP

– percentage predicted of force expiratory

volume in 1 second.

The difference in wheeze occupation rate

between both studied groups is presented in Table 3.

A superior reduction in occupation rate in both

inspiratory and expiratory respiratory phases (figure

2), after physiotherapy treatment was observed,

reaching statistical significance for the inspiratory

phase (p=0.019).

Figure 2: Wheezes occupation rate differences between

control and physiotherapy groups. Significance level set to

*p<0.05.

4 DISCUSSION

The combination between different algorithms

contributes to establish objective measures to assess

the effect of respiratory physiotherapy in patients

with acute respiratory diseases (e.g., LRTI).

The values collected at baseline for FEV1

(67.7

± 28.2) and FVC

(69.8 ± 23.9) were lower than

those previously reported for patients with LRTI.

Melbye et al. (1994) found that FEV1

in patients

with upper or lower respiratory tract infection was

90% of the predicted value, however various factors

could affect this value, such as cough, dyspnoea and

smoking habits. Such factors were highly prevalent

in the present sample. Furthermore, in the study of

Melbye et al. (1994) the spirometry test was

performed in the standing position while in the

present study patients presented with severe

symptoms of pain, cough and dyspnoea and were

instructed to perform the test in the sitting position,

0,00

0,02

0,04

0,06

0,08

0,10

0,12

Wheeze Occuptaion Rate

Control Group Experimental Group

I - Pre

p = 0.061

p = 0.019

E-Pre

I - Post

E-Post

HEALTHINF2013-InternationalConferenceonHealthInformatics

236

Table 2: Paired sample t-test results for pre-post treatment analysis on control and experimental groups.

Control Group Experimental Group

Pre Post p t Pre Post p t

(%)

0.111 ± 0.148 0.022 ± 0.062

< 0.001 5.462 0.092 ± 0.141 0.004 ± 0.019 < 0.001 4.777

(%)

0.113 ± 0.132 0.041 ± 0.077

< 0.001 4.392 0.105 ± 0.153 0.019 ± 0.054 < 0.001 4.402

(Hz)

241.3 ± 60.1 415.5 ± 201.1 0.195 -1.554 360.3 ± 221.1 1402 ± 1531 0.555 -0.841

(Hz)

221.2 ± 85.6 396.8 ± 208.1 0.243 -1.368 423.2 ± 168.6 432.8 ± 269.1 0.915 -0.111

(%)

0.050 ± 0.111 0.032 ± 0.074 0.374 1.000 0.083 ± 0.117 0.000 ± 0.000 0.500 1.000

(%)

0.107 ± 0.220 0.106 ± 0.301 0.999 0.001 0.238 ± 0.224 0.142 ± 0.377 0.652 0.475

R - wheeze occupation rate; F - main frequency; T - type of wheeze. Subscript I and E stand for inspiration and expiration, respectively.

Significance level set to p<0.05.

which could also have affected the test performance.

Table 3: Paired sample t-test results for the two groups.

CG EG p t

I - Pre

0.111 ± 0.148 0.092 ± 0.141 0.455 0.618

E - Pre

0.113 ± 0.132 0.105 ± 0.153 0.749 0.282

I - Post

0.022 ± 0.062 0.004 ± 0.019

0.019 2.762

E - Post

0.041 ± 0.077 0.019 ± 0.054 0.061 1.907

CG - Control group; EG - experimental group; R - wheeze

occupation rate; Subscript I and E stand for inspiration and

expiration, respectively. Significance level set to p<0.05.

There were no significant differences in

inspiratory and expiratory wheezes occupation rates

pre-treatment. This shows that both studied groups

were similar in terms of wheezes parameters at

baseline assessment.

The results of pre/post treatment analysis (table

2) showed significant statistical decrease in both

inspiratory and expiratory wheeze occupation rate

for control and experimental groups. This was an

expected outcome, because both groups received, at

least, standard medication treatment i.e., antibiotics.

The experimental group, which received respiratory

physiotherapy, presented a significantly lower

inspiratory wheeze occupation rate (p=0.019) and a

pattern of decreased expiratory wheeze occupation

rate (p=0.061). As previously stated by Sovijärvi et

al., (2000), more severe cases of respiratory

infection can also present wheezes in the inspiratory

phase. The sharp decrease on inspiratory wheeze

occupation rate seems to suggest that the respiratory

physiotherapy plays an important role on patients

with more severe conditions. Another result that

supports this theory is the non-existence of

inspiratory polyphonic wheezes post-treatment in the

experimental group, and, although not statistically

significant, a sharp decrease in the expiratory

polyphonic wheezes.

5 CONCLUSIONS

This study suggests that by combining respiratory

physiotherapy with the standard medical therapy

more effective results in the reduction of respiratory

wheeze can be achieved in patients with LRTI.

Furthermore, the use of wheeze and respiratory

phase detectors appears to be a responsive measure

to evaluate the efficacy of treatments in LRTI.

Further research to assess responsiveness with a

larger sample is nevertheless needed.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge the funding

provided to this project, “Sounds4Health”, by

Quadro de Referência Estratégico Nacional

(QREN), on a partnership between University of

Aveiro and ISA (Intelligence Sensing Anywhere).

The authors would also like to thank to

physicians at Hospital Infante D. Pedro for their help

during the recruiting phase of the study and to all

participants.

REFERENCES

American Association of Cardiovascular and Pulmonary

Rehabilitation, 2006. Guidelines for pulmonary

rehabilitation programs, USA, Human Kinetics.

American College of Sports Medicine, 2006. Guidelines

for Exercise Testing and Prescription, Philadelphia,

Lippincott Williams and Wilkins.

Baughman, R. P. & Loudon, R. G., 1984. Quantitation of

wheezing in acute asthma. Chest, 86, 718-22.

British Thoracic Society, 2001. Pulmonary rehabilitation.

Thorax, 56, 827-34.

Charbonneau, G., Ademovic, E., Cheetham, B.,

Malmberg, L., Vanderschoot, J. & Sovijärvi, A., 2000.

AutomaticWheezeandRespiratoryPhaseDetectorstoEvaluateRespiratoryPhysiotherapyinLRTI-APreliminaryStudy

237

Basic techniques for respiratory sound analysis. Eur

Respir Rev., 10, 625-635.

Earis, J. & Cheetham, B., 2000. Future perspectives for

respiratory sound research. Eur Respir Rev, 10, 641-

646.

Gabor, D. 1946. Theory of communication. Journal of the

IEEE, 93, 429-457.

Garrod, R. & Lasserson, T., 2007. Role of physiotherapy

in the management of chronic lung diseases: an

overview of systematic reviews. Respiratory

Medicine, 101, 2429-2436.

Graham, D., 2008. The patient with possible pneumonia.

The Foundation Years, 4, 14-19.

Huq, S. & Moussavi, Z., 2010. Automatic Breath Phase

Detection Using Only Tracheal Sounds. 2010 Annual

International Conference of the Ieee Engineering in

Medicine and Biology Society (Embc), 272-275.

Laennec, R., 1935. A treatise on the diseases of the chest

and mediate auscultation, New York, Samuel Wood

and Sons.

Melbye, H., Kongerud, J. & Vorland, L., 1994. Reversible

airflow limitation in adults with respiratory infection.

Eur Respir J, 7, 1239-45.

Meslier, N., Charbonneau, G. & Racineux, J. L., 1995.

Wheezes. European respiratory journal, 8, 1942-8.

Miller, M. R. 2005. Standardisation of spirometry.

European Respiratory Journal, 26, 319-338.

Moher, D., Hopewell, S., Schulz, K. F., Montori, V.,

Gotzsche, P. C., Devereaux, P. J., Elbourne, D., Egger,

M. & Altman, D. G., 2010. CONSORT 2010

explanation and elaboration: updated guidelines for

reporting parallel group randomised trials. BMJ, 340,

c869.

Oliveira, D., Pinho, C., Marques, A. & Dinis, J., 2011.

Validation of a time-frequency wheeze detector in

cystic fibrosis: A pilot study. Eur Respirator J, 38,

92s.

Piirilä, P., Sovijärvi, A., Earis, J., Rossi, M., Dalmasso, F.,

Stoneman, S. & Vanderschoot, J., 2000. Reporting

results of respiratory sound analysis. European

Respiratory Review, 10, 636-640.

Postiaux, G., 2004. Fisioterapia respiratória pediátrica, o

tratamento guiado por ausculta pulmonar.

Pryor, J. & Prasad, S., 2008. Physiotherapy for

Respiratory and Cardiac Problems: Adults and

Paediatrics, UK, Churchill Livingstone.

Qiu, Y., Whittaker, A. R., Lucas, M. & Anderson, K.,

2005. Automatic wheeze detection based on auditory

modelling. Proceedings of the Institution of

Mechanical Engineers Part H-Journal of Engineering

in Medicine, 219, 219-227.

Sovijärvi, A., Malmberg, L., Charbonneau, G.,

Vanderschoot, J., Dalmasso, F., Sacco, C., Rossi, M.

& Earis, J. E., 2000. Characteristics of breath sounds

and adventitious respiratory sounds. Eur Respir Rev,

10, 591–596.

Taplidou, S. A. & Hadjileontiadis, L. J., 2007. Wheeze

detection based on time-frequency analysis of breath

sounds. Computers in Biology and Medicine, 37,

1073-1083.

Ward, D. & Ayres, J., 2000. Pneumonia and acute

bronchitis. The European respiratory journal

Supplement, 5, 105-127.

Waris, M., Helisto, P., Haltsonen, S., Saarinen, A. &

Sovijarvi, A. R., 1998. A new method for automatic

wheeze detection. Technol Health Care, 6, 33-40.

Weiner, P., Man, A., Weiner, M., Rabner, M., Waizman,

J., Magadle, R., Zamir, D. & Greiff, Y., 1997. The

effect of incentive spirometry and inspiratory muscle

training on pulmonary function after lung resection. J

Thorac Cardiovasc Surg, 113, 552-7.

Woodhead, M., Blasi, F., Ewig, S., Garau, J., Huchon, G.,

Ieven, M., Ortqvist, A., Schaberg, T., Torres, A., van

der Heijden, G., Read, R., Verheij, T. J. M., Soc, E. R.

& Microbiology, E. S. C., 2011. Guidelines for the

management of adult lower respiratory tract infections

- Summary. Clinical Microbiology and Infection, 17,

1-24.

Yildirim, I., Ansari, R. & Moussavi, Z., 2008. Automated

Respiratory Phase and Onset Detection Using Only

Chest Sound Signal. 2008 30th Annual International

Conference of the IEEE Engineering in Medicine and

Biology Society, Vols 1-8, 2578-2581.

HEALTHINF2013-InternationalConferenceonHealthInformatics

238