Characterization and cross-comparison of digital stethoscopes for telehealth remote patient auscultation

Publications Home » Browse Publications » Characterization and cross-comparison of digital stethoscopes for telehealth remote patient auscultation

About Me

Yuriy Ilkovych

Pulmonology

Posted by Yuriy Ilkovych -

Jan 22 -

824 Views 0 Comments 0 Likes 1 Review

Expert Rating:

Highlights

•
Digital Stethoscopes are becoming increasingly common and are enablers of digital medicine.
•
An objective cross-evaluation of the performance of digital stethoscopes is presented.
•
New configuration of stethoscopes is evaluated against existing standards.
•
Wearable digital stethoscopes are found to be beneficial in remote patient monitoring.

Abstract

— Stethoscopes are commonly used to diagnose cardiac and respiratory diseases. The advent of digital stethoscopes added the capability to remotely monitor the cardiorespiratory well-being of patients and facilitate digital remote auscultations for physicians to examine their patients outside of the clinic. To increase the adoption of a hybrid diagnosis between in-person and remote auscultations, it is important to characterize the frequency response of different configurations of digital stethoscopes. A standard testing platform was adopted from the literature and re-validated by seven physicians. Two wearable and two non-wearable digital stethoscopes were investigated. A similarity factor score was used to analyze select time-synchronized acoustic events that achieved high fidelity between stethoscopes. The frequency responses between devices were compared using the correlation coefficient. Two devices reported the highest correlation coefficient of 0.72 and 0.75 in the wearable and non-wearable categories, respectively. The correlation coefficient of the frequency responses between the same pair of devices on the human chest was 0.91. With such a high agreement in the frequency response between the wearable and non-wearable digital stethoscope, it can be concluded that the configuration of digital stethoscope is not a limiting factor of performance, and both are acceptable and desirable for physicians to adopt either configuration in remote delivery of healthcare.

Keywords

Digital stethoscope

Telehealth

Auscultation

Frequency response

Similarity factor

1. Introduction

Cardiac and respiratory diseases such as coronary heart disease, heart valve disease, chronic obstructive pulmonary disease (COPD), and asthma cause heavy societal and economic burdens [[1], [2], [3], [4], [5]]. These diseases are often first detected by physicians using a stethoscope upon patient complaints. The traditional stethoscope usually consists of a hollow tube attached to a chest piece made from a wider-based diaphragm coupled cavity and a smaller hollow, open bell. Heart and lung sounds vibrate the diaphragm in contact with the patient's chest wall and transmit as sound waves through the cavity and up the hollow tube towards the earpiece. Although most physicians are trained to accurately diagnose diseases related to the cardiovascular and respiratory systems through a single auscultation event, there exist chances of misdiagnosis or underdiagnosis due to potential limitations such as the absence of symptoms during an examination or even presbycusis of the physician. Therefore, this has called for the advent of digital stethoscopes, which allow the recording of subtle anomalies present in heart and lung sounds [6]. The recording can be used for in-depth analysis among physicians and researchers to advance optimal intervention strategies for difficult patient conditions. Other advantages of the digital stethoscope include the ability to amplify the chest sound for optimal listening, the potential to be integrated with computerized algorithms for automated anomaly detection, and the potential to be used as a remote monitoring device for better patient management [6,7].

Hand-held auscultation using a digital stethoscope such as the commonly used 3M™ Littmann® Electronic Stethoscopes Model 3200 is susceptible to motion artifacts due to the device's rigidity and weight and/or the accidental rubbing or jittery hands of users throughout recordings [8]. For this reason, there has been an increasing demand for the design of lighter portable, wearable digital stethoscopes in recent years [[8], [9], [10], [11]]. With that, motion artifacts can also be reduced, especially with a wearable stethoscope. However, to ensure the reliability of digital stethoscopes in diagnostic care, the acoustic functionality of these newer configurations of stethoscopes must be assessed. For example, the device must have an acceptable acoustic sensitivity to normal and abnormal heart and lung sounds. That led to the design and development of a standardized testing platform for the acoustic characterization of digital stethoscopes [12]. Moreover, previous works have established the importance of having a standard sound source (preferably an auscultation recording over a white noise generator at different frequency levels) and a standard objective performance metric to evaluate the acoustic caliber of the digital stethoscopes [13].

The selections of these stethoscopes for comparison are based on pragmatic criteria. This study aims to objectively qualify the performance of different stethoscopes by cross-comparison with each other. This is alike medical device validation, where a gold standard is usually used as the reference. However, as digital stethoscopes are not binary diagnostic devices but diagnostic tools, a gold standard does not exist. Therefore, we cross-compare the performance of different devices and the most clinically used model. The devices are also chosen to represent the different configurations available for digital stethoscopes. Evaluation of digital stethoscopes of different configurations meant for different use environments and use cases is important for the move towards distributed healthcare where digital stethoscopes are not only to be used by trained professionals but also patients themselves for home healthcare and telehealth.

The pragmatic criteria are as follows:

•
All the stethoscopes must have at least one regulatory clearance.
•
Digitalized bioacoustics recorded must be extractable for performance evaluation.
•
Sampling frequency of the stethoscope must be at least 4 kHz.
•
All stethoscopes are of different configurations.

In this study, we compared the acoustic characteristics of four digital stethoscopes with different dimensions and configurations: (a) A miniature wearable digital stethoscope – Device 1, (b) A hybrid digital stethoscope with a dual-head chest piece and a tube connecting a binaural earpiece – Device 2, (c) A digital stethoscope with a single-head chest piece and a tube connecting a binaural earpiece – Device 3, and (d) A wearable digital stethoscope - Device 4. All selected stethoscopes have at least one regulatory clearance as a medical device. The main aim of the stethoscope comparison study is to provide insights into digital stethoscopes' suitability for generalized clinical applications based on their sensitivity in particular frequency ranges. Concretely, the study utilizes a standard benchtop testing platform [12] and objective metrics developed from a previous study [13] to ensure that only high-fidelity signals of interest among all the time synchronized recordings obtained from the devices were adopted in the present study. Furthermore, the relative correlation coefficients in the frequency response curves between stethoscopes’ recordings were also computed. Lastly, one of the two wearable digital stethoscopes that showed the higher average relative correlation coefficient and one hand-held digital stethoscope were evaluated on the human chest.

2. Methods

2.1. Design of the chest phantom model

In recent years, chest phantom models have been widely adopted for evaluating the acoustic properties of different stethoscopes; the sound quality played through the phantom model is reproducible and therefore used as a standardized method for stethoscope evaluation [12,14,15]. In this regard, we constructed a similar chest phantom model as an objective testing platform for comparing the acoustic characteristics of different stethoscopes, as shown in Fig. 1. The details on the materials are available in the literature [12]. The speaker used in this phantom model had a frequency response of 50 Hz to 18 kHz. However, because this is a comparative study between test devices, the frequency profile of the speaker is not of interest.

2.2. Validation of the chest phantom model

Seven physicians were recruited to validate the chest phantom model before adopting it for use in this stethoscope comparison study. Specifically, three physicians were involved in an in-person auscultation validation, while another four participated in an online auscultation test. The test uses normal and abnormal breath samples from an online database as the auscultation sounds played through the chest phantom model [16] for physicians to auscultate with their traditional stethoscopes in person or through an electronic stethoscope recording remotely.

The auscultation location was marked directly on top of the speaker. For the in-person auscultation study, each physician was instructed to bring their acoustic stethoscopes to auscultate the chest phantom model by hand on a marked position in the middle of the phantom model surface. The physicians were free to apply pressure as they would normally do when auscultating a patient. Normal and wheezy breath sounds were played, one at a time, at different volumes (for the adjustment of audibility). At each volume level, the physicians were asked whether the auscultation was similar to their clinical auscultation experience. All three physicians must agree that the normal breath and wheeze sounds played through the chest phantom model give a similar clinical auscultation experience for the model to be qualified.

The optimal sound volume that all three physicians acknowledged is the suitable loudness they would expect while performing the manual auscultation was noted and used for re-recording two sets of normal breath and two sets of wheeze sounds in the remote validation test. The re-recordings were performed using a digital stethoscope (3 M™ Littmann® Electronic Stethoscopes Model 3200). The auscultation recording was made on the same marked position in the middle of the phantom model surface and gently depressed on the surface using a light tensioning rod. As the digital stethoscope used for re-recording can be used as a traditional stethoscope while recording, a medic auscultated the model to determine suitability of pressure before recording.

The re-recorded sounds with their corresponding original sounds were sent to four physicians for a secondary validation test. Based on each pair of sounds given, the physicians were to inform whether the difference in the sounds between the original recorded from the anterior chest and the re-recorded version from the chest phantom model would affect their judgment in diagnosis (i.e., classify abnormal lung sound from normal lung sound) if they were to use the re-recordings as means of remote auscultation. Overall, there were 16 responses (four physicians’ responses to the four sets of recordings), and to qualify the chest model in this secondary validation test, at least 80% of the responses from the seven physicians must state that the re-recordings would not affect their judgment in diagnosis.

2.3. Test device descriptions

Table 1 summarizes the technical specifications of the digital stethoscopes in the present study. All devices had obtained at least 1 regulatory clearance globally when the manuscript was written.

Table 1. Summary of technical specifications of the digital stethoscopes.

Empty Cell	Device 1	Device 2	Device 3	Device 4
Model name (Manufacturer)	AeviceMD (Aevice Health, Singapore)	CORE - Digital Stethoscope (3M™ Littmann®, USA)	Model 3200 (3M™ Littmann®, USA)	Stemoscope (Hulu Devices, China)
Frequency range (Hz)	20–2000	20–2000	20–2000	20–4000
Diaphragm diameter, ⌀ (mm)	15	33 (for the bell chest piece)	51	38
Configuration	Miniature tubeless digital stethoscope that can be either hand-held or attached to the chest using adhesive tape.	Hand-held hybrid∗ digital stethoscopes with a binaural tube, dual-head bell, and diaphragm chest piece. ∗Hybrid: There are manual and digital interchangeable configurations.	Hand-held digital stethoscope with a binaural tube and a single-head diaphragm chest piece.	Tubeless digital stethoscope. Can be used as a hand-held or a wearable stethoscope with adhesive tape.

2.4. Comparison of the acoustic characteristic of four digital stethoscopes

The acoustic characteristics of the four digital stethoscopes were first evaluated on the chest phantom model and next on a human chest.

2.4.1. Evaluation of the chest auscultation playbacks from the chest phantom model

Four stethoscopes were placed equidistant from the speaker on the top of the chest phantom model. The test signal was a continuous looping 5-s segment of pre-recorded heart sound from a resting subject with a heart rate of 50 bpm with an impulse appended to the front of the test signal for time synchronization across the recording from all the test devices. The recording was repeated by changing each stethoscope's position clockwise. The ambient sound pressure level was maintained between 30 and 50 dB.

In each recording, the acoustic events of the S1 sounds were extracted. Each acoustic event had 2048 samples, about 0.5 s. The similarity coefficient, Equation. (1) and the similarity factor, Equation. (2) adapted from Ref. [13] were used to select the acoustic events that had high fidelity (i.e., similarity factor higher than 0.70) for the frequency response evaluation. The relative similarities in the frequency response curves between devices were computed using the correlation coefficient,(1)

S m n = (\sum i = 1 N x m i x n i) 2 (\sum i = 1 N (x m i) 2) (\sum i = 1 N (x n i) 2) ">

where

x m i ">

and

x n i ">

denotes the

i ">

th sample of signals

m ">

and

n ">

, respectively. The

m ">

and

n ">

can be acoustic events extracted from the same or different device(s).

N ">

is the number of samples in each acoustic event, which in our study is 2048 samples. The similarity factor,

S m ">

, was computed by normalizing the similarity coefficient with the maximum similarity coefficients across all acoustic events,(2)

S m = \sum i = 1 M s m i max j = 1 M (\sum i = 1 M s j i) ">

where M denotes the total number of acoustic events extracted from the recording.

2.4.2. Evaluation of the chest auscultation on the human chest

Although the chest phantom model could provide a standardized sound source for the acoustic characterization of digital stethoscopes, it is still imperative to evaluate actual chest sounds. Therefore, one of the two wearable digital stethoscopes and one non-wearable digital stethoscope that had achieved the higher average correlation coefficient scores were chosen for the human chest evaluation. Fig. 2 depicts the placement of two devices for two rounds of recordings. Device 1 and 2 placements were swapped in the second round of recording to ensure sensitivity fairness due to proximity from the sound source. Each recording was 15 s long. An impulse signal was generated by tapping the subject's sternum as the reference point for time synchronization.

3. Results and discussion

3.1. Validation of the chest phantom model

Overall, three physicians who have performed auscultation on the chest phantom model with their acoustic stethoscopes unanimously agreed that the auscultation on the chest phantom model was close to their clinical experience. In addition, the optimal volume at which the auscultation became visible and closest to the manual experience was also taken as the volume played throughout the remaining study.

In the secondary validation, there were 15 responses out of 16 total responses (i.e., 94%) from the physicians that noticed no significant difference between original and re-recorded sounds, and all 16 responses (i.e., 100%) informed that the differences did not affect their judgment in diagnosis if the sound samples were to be used as means of remote auscultation. Here, only feedback concerning the deterioration in sound quality, when played through the chest phantom model, was regarded as a negative response. Otherwise, the feedback concerning sound volume was disregarded as it can be resolved by manually setting it to the physician's comfort level.

3.2. Comparison of the acoustic characteristic of four digital stethoscopes

3.2.1. Evaluation of the chest auscultation playbacks in the chest phantom model

Overall, 11 out of 16 acoustic events (i.e., 81%) achieved a similarity factor greater than 0.70. Fig. 3 shows one example of the acoustic events extracted from the time synchronization method using the impulsive signal as a reference point. The mean frequency responses across the acoustic events measured by each digital stethoscope were compared in Fig. 4(a). A zoom-in plot for frequencies between 0 and 300 Hz is depicted in Fig. 4(b). Table 2 summarizes the device(s) with the highest and lowest spectral power in different frequency ranges observed in Fig. 4(a) and their respective clinical relevance. Physicians may take reference on which digital stethoscope is more suited for the pathologies they intend to diagnose. For example, Device 2 can be recommended for auscultation for bioacoustics relevant to stenosis due to its highest spectral power in the frequency range of less than 100 Hz compared to the devices. Device 1 can be recommended for high-pitched rhonchus. Device 4 shows consistently high spectral power in the frequency range of 100–1000 Hz and can be opted for auscultation of crackles, wheezes, and murmurs within that frequency range.

Table 2. Summary of the device(s) with the highest and lowest spectral power in different frequency ranges and their respective clinical relevance.

Frequency	Clinical relevance [[20], [21], [22], [23], [24], [25], [26]]	Device(s) with highest spectral power
<100 Hz	- Normal heart sound - Stenosis	Device 2
100–150 Hz	- Normal heart and lung sounds - Crackle - Wheeze	Devices 1, 2, and 4
150–200 Hz	- Crackle - Wheeze	Device 4
200–300 Hz	- Murmurs - Crackle - Wheeze	Devices 1 and 4
300–400 Hz	- Murmurs - Crackle - Rhonchus	Device 4
400–750 Hz	- Murmurs - Crackle - Rhonchus	Devices 1, 2, and 4
750–1000 Hz	- Rhonchus	Devices 1 and 4
>1000 Hz	- Rhonchus	Device

Table 3 shows that Device 2 achieved the highest average correlation coefficient (i.e., 0.75) in the frequency response curves compared with other devices, followed by Device 1, which reported the second highest average correlation coefficient score (i.e., 0.72).

Table 3. The correlation coefficient in the frequency response curves between signals collected from different devices.

Empty Cell	Original	Device 1	Device 2	Device 3	Device 4
Original	1.00	0.61	0.73	0.64	0.59
Device 1	0.61	1.00	0.88	0.22	0.91
Device 2	0.73	0.88	1.00	0.32	0.81
Device 3	0.64	0.22	0.32	1.00	0.14
Device 4	0.59	0.91	0.81	0.14	1.00
Average correlation score	0.71	0.72	0.75	0.46	0.69

3.2.2. Evaluation of chest auscultation on the human chest

Fig. 5 shows one example of the acoustic events from chest auscultation extracted from the time synchronization method using the impulsive signal as a reference point. Fig. 6(a) compares the mean frequency responses from the chest auscultation across the acoustic events measured by each digital stethoscope. A zoom-in plot for frequencies between 0 and 300 Hz is depicted in Fig. 6(b). The correlation coefficient in the frequency response curves between Device 1 and Device 2 was 0.91.

Device 1 and 2 has a peak response at 31 Hz and 35 Hz, respectively. Device 1 had a higher sensitivity of up to 3.8 dB than Device 2, between 50 Hz and 60 Hz. The discrepancies increased from 3.8 to 8.8 dB between the 60 Hz and 200 Hz frequency bands. The greatest discrepancy was 57 dB, observed at around 900 Hz. Hence, Device 1 is more sensitive to collecting sounds of the higher frequency profile, as shown in Fig. 6(a) and (b).

3.3. Strengths, limitations, and future work

The present study employed an objective method to compare the acoustic characteristics of different stethoscope models by adopting both the standard testing platform [12] and the standard metric (i.e., the similarity factor) [13] to segment acoustic events that share high fidelity in recordings obtained across all stethoscopes. Before the evaluation, we added an in-person and online validation of the sound quality from the chest phantom model by seven physicians. Similar subjective assessment by physicians used for the validation of the chest phantom model was also used by other studies to evaluate the agreement between physicians in assessing different cardiac and lung sound anomalies using different stethoscope models [[17], [18], [19]].

A higher sensitivity toward the higher frequency band is a double-edged sword feature. This is because abnormal lung sounds, such as high-pitched wheezes and coughs, and undesired acoustics signal from subjects’ environmental speeches all span the higher frequency range [20,[27], [28], [29]]. This calls for future work integrating smart algorithms to distinguish the differences between high-frequency abnormal lung sounds and high-frequency noises. With that, a smart digital stethoscope that can automatically detect anomalies in cardiorespiratory health will be feasible in the market in the future. Furthermore, although we presented a correlation coefficient in the frequency response between Device 2 and Device 1, a clinical study to evaluate the agreement in diagnosis outcome between the two and also when compared to in-person auscultation with a traditional stethoscope is desirable to increase the clinical confidence in adopting digital stethoscopes in clinical practice.

4. Conclusion

Chest auscultation is the most basic diagnostic method used in hospitals and clinics. Physicians are trained to diagnose cardiac and respiratory anomalies using the acoustic stethoscope. However, in some subtle cases, the use of acoustic stethoscopes, compared to electronic stethoscopes, is limited to amplification, frequency band selection, further evaluation, and discussion with other, more experienced physicians to advance the understanding of different pathologies. For new digital stethoscopes to be more widely adopted by physicians, the acoustic characteristics should be similar to the commonly used digital stethoscopes. Hence, this study presented an experimental and analytical framework to evaluate such properties and does so on some emerging electronic stethoscope variants in the market.

Author contribution

Yi Yang Ang: Conceptualization, Investigation, Validation, Formal analysis, Writing – review & editing. Liren Aw: Investigation, Writing – original draft, Formal analysis. Vivian Koh: Investigation, Writing – original draft, Formal analysis. Rex Tan: Conceptualization, Methodology, Validation, Formal analysis, Writing – review & editing, Supervision.

Funding sources

This research did not receive any specific grant from public, commercial, or not-for-profit funding agencies.

Declaration of competing interest

All authors in the paper are employees of Aevice Health Pte Ltd, a medical device manufacturer. However, this research is not of commercial value and did not receive any specific grant from public, commercial, or not-for-profit funding agencies.

References

[1]

V. Cukic, V. Lovre, D. Dragisic, A. Ustamujic

Asthma and chronic obstructive pulmonary disease (COPD)–differences and similarities

Mater Sociomed, 24 (2) (2012), p. 100

Crossref Google Scholar
[2]

CDC

Vital signs: asthma prevalence, disease characteristics, and self-management education: United States, 2001--2009

MMWR Morb Mortal Wkly Rep, 60 (17) (2011), pp. 547-552

Google Scholar
[3]

T.L. Petty, D.E. Doherty

The national lung health education program: roots, mission, future directions

Respir Care, 49 (6) (2004), pp. 678-683

View in Scopus Google Scholar
[4]

J. Leal, R. Luengo-Fernández, A. Gray, S. Petersen, M. Rayner

Economic burden of cardiovascular diseases in the enlarged European Union

Eur Heart J, 27 (13) (2006), pp. 1610-1619

Crossref View in Scopus Google Scholar
[5]

C. Cook, G. Cole, P. Asaria, R. Jabbour, D.P. Francis

The annual global economic burden of heart failure

Int J Cardiol, 171 (3) (2014), pp. 368-376

View PDF View article View in Scopus Google Scholar
[6]

S. Swarup, A.N. Makaryus

Digital stethoscope: Technology update

Med devices (Auckland, NZ), 11 (2018), p. 29

Crossref View in Scopus Google Scholar
[7]

S. Leng, R.S. Tan, K.T.C. Chai, C. Wang, D. Ghista, L. Zhong

The electronic stethoscope

Biomed Eng Online, 14 (1) (2015), pp. 1-37

Google Scholar
[8]

S.H. Lee, et al.

Fully portable continuous real-time auscultation with a soft wearable stethoscope designed for automated disease diagnosis

Sci Adv, 8 (21) (2022), p. eabo5867

View in Scopus Google Scholar
[9]

G. Yilmaz, et al.

A wearable stethoscope for long-term ambulatory respiratory health monitoring

Sensors, 20 (18) (2020), p. 5124

Crossref Google Scholar
[10]

S.H. Lee, Y. Kim, W. Yeo

Advances in microsensors and wearable bioelectronics for digital stethoscopes in health monitoring and disease diagnosis

Adv Healthcare Mater, 10 (22) (2021), Article 2101400

View in Scopus Google Scholar
[11]

V.C.A. Koh, R.X. Tan, Y.Y. Ang

A miniature wearable microphone sensor for cardiopulmonary monitoring

2020 IEEE sensors (2020), pp. 1-4

Crossref Google Scholar
[12]

D. Weiss, et al.

An in vitro acoustic analysis and comparison of popular stethoscopes

Med Dev Evid Res (2019), pp. 41-52

Crossref View in Scopus Google Scholar
[13]

L.J. Nowak, K.M. Nowak

Acoustic characterization of stethoscopes using auscultation sounds as test signals

J Acoust Soc Am, 141 (3) (2017), pp. 1940-1946

Crossref View in Scopus Google Scholar
[14]

R.L. Watrous, D.M. Grove, D.L. Bowen

Methods and results in characterizing electronic stethoscopes

Computers in cardiology (2002), pp. 653-656

View in Scopus Google Scholar
[15]

V. Rennoll, I. McLane, D. Emmanouilidou, J. West, M. Elhilali

Electronic stethoscope filtering mimics the perceived sound characteristics of acoustic stethoscope

IEEE J Biomed Heal informatics, 25 (5) (2020), pp. 1542-1549

Google Scholar
[16]

B.M. Rocha, et al.

Α respiratory sound database for the development of automated classification

Thessaloniki, Greece

Precision medicine powered by pHealth and connected health: ICBHI 2017 (2018), pp. 33-37

18-21 November 2017

Crossref View in Scopus Google Scholar
[17]

M. Fan, et al.

Real-world evaluation of the Stemoscope electronic tele-auscultation system

Biomed Eng Online, 21 (1) (2022), p. 63

View in Scopus Google Scholar
[18]

O. Haskel, E. Itelman, E. Zilber, G. Barkai, G. Segal

Remote auscultation of heart and lungs as an acceptable alternative to legacy measures in quarantined COVID-19 patients—prospective evaluation of 250 examinations

Sensors, 22 (9) (2022), p. 3165

Crossref View in Scopus Google Scholar
[19]

T. Hirosawa, Y. Harada, K. Ikenoya, S. Kakimoto, Y. Aizawa, T. Shimizu

The utility of real-time remote auscultation using a Bluetooth-connected electronic stethoscope: open-label randomized controlled pilot trial

JMIR mHealth uHealth, 9 (7) (2021), Article e23109

Crossref View in Scopus Google Scholar
[20]

S. Reichert, R. Gass, C. Brandt, E. Andrès

Analysis of respiratory sounds: state of the art

Clin Med Circ Respir Pulm Med, 2 (2008), Article CCRPM-S530

Google Scholar
[21]

N. Gavriely, M. Nissan, A.H. Rubin, D.W. Cugell

Spectral characteristics of chest wall breath sounds in normal subjects

Thorax, 50 (12) (1995), pp. 1292-1300

Crossref View in Scopus Google Scholar
[22]

P.J. Arnott, G.W. Pfeiffer, M.E. Tavel

Spectral analysis of heart sounds: relationships between some physical characteristics and frequency spectra of first and second heart sounds in normals and hypertensives

J Biomed Eng, 6 (2) (1984), pp. 121-128

View PDF View article View in Scopus Google Scholar
[23]

M.E. Tavel

Clinical phonocardiography and external pulse recording

Clinical phonocardiography and external pulse recording (1972), p. 322

Google Scholar
[24]

N. Hayashi

Detection of pneumothorax visualized by computer analysis of bilateral respiratory sounds

Yonago Acta Med, 54 (4) (2011), p. 75

View in Scopus Google Scholar
[25]

M. Munakata, et al.

Spectral and waveform characteristics of fine and coarse crackles

Thorax, 46 (9) (1991), pp. 651-657

Crossref View in Scopus Google Scholar
[26]

D. Callahan, J. Waugh, G.A. Mathew, W.M. Granger

Stethoscopes: what are we hearing?

Biomed Instrum Technol, 41 (4) (2007), pp. 318-323

View in Scopus Google Scholar
[27]

M.A. Picheny, N.I. Durlach, L.D. Braida

Speaking clearly for the hard of hearing II: acoustic characteristics of clear and conversational speech

J Speech Lang Hear Res, 29 (4) (1986), pp. 434-446

View at publisherCrossref View in Scopus Google Scholar
[28]

H. Erdogan, J.R. Hershey, S. Watanabe, J. Le Roux

Deep recurrent networks for separation and recognition of single-channel speech in nonstationary background audio

New Era Robust Speech Recognit Exploit Deep Learn (2017), pp. 165-186

View at publisherCrossref Google Scholar
[29]

B.-S. Lin, B.-S. Lin

Automatic wheezing detection using speech recognition technique

J Med Biol Eng, 36 (2016), pp. 545-554

Expert Reviews

fdgdfg

by Yuriy Ilkovych |

Pros:

Cons:

Description: fdgdfgdfgdfgdfgdfg

Recommended

Continue Reading »

Compare