Spectral and waveform characteristics of fine and coarse crackles

Advances in Wearable Stethoscope Technology: Opportunities for the Early Detection and Prevention of Cardiovascular Diseases

Wearable technology, including devices like Apple and Samsung watches, Fitbits, and smart rings, has become widely popular. However, while these consumer electronics are readily available, they do not yet meet the accuracy and safety standards required for medical devices by the U.S. Food and Drug Administration (FDA). The COVID-19 pandemic has spurred demand for wearable medical devices, particularly those that can support telemedicine and telehealth. Among these, wearable electronic stethoscopes hold significant promise for early detection and prevention of cardiovascular diseases, which remain the leading cause of death globally. This review highlights the potential of wearable electronic stethoscopes to transform cardiovascular health management by enabling early diagnosis and self-monitoring. Additionally, it examines the current challenges and technological advancements needed to overcome them, underscoring the vital role that wearable electronic stethoscopes could play in improving global health outcomes.

Deep learning-based lung sound analysis for intelligent stethoscope

Auscultation is crucial for the diagnosis of respiratory system diseases. However, traditional stethoscopes have inherent limitations, such as inter-listener variability and subjectivity, and they cannot record respiratory sounds for offline/retrospective diagnosis or remote prescriptions in telemedicine. The emergence of digital stethoscopes has overcome these limitations by allowing physicians to store and share respiratory sounds for consultation and education. On this basis, machine learning, particularly deep learning, enables the fully-automatic analysis of lung sounds that may pave the way for intelligent stethoscopes. This review thus aims to provide a comprehensive overview of deep learning algorithms used for lung sound analysis to emphasize the significance of artificial intelligence (AI) in this field. We focus on each component of deep learning-based lung sound analysis systems, including the task categories, public datasets, denoising methods, and, most importantly, existing deep learning methods, i.e., the state-of-the-art approaches to convert lung sounds into two-dimensional (2D) spectrograms and use convolutional neural networks for the end-to-end recognition of respiratory diseases or abnormal lung sounds. Additionally, this review highlights current challenges in this field, including the variety of devices, noise sensitivity, and poor interpretability of deep models. To address the poor reproducibility and variety of deep learning in this field, this review also provides a scalable and flexible open-source framework that aims to standardize the algorithmic workflow and provide a solid basis for replication and future extension: https://github.com/contactless-healthcare/Deep-Learning-for-Lung-Sound-Analysis .

Regularity and mechanism of fake crackle noise in an electronic stethoscope

Background: Electronic stethoscopes are widely used for cardiopulmonary auscultation; their audio recordings are used for the intelligent recognition of cardiopulmonary sounds. However, they generate noise similar to a crackle during use, significantly interfering with clinical diagnosis. This paper will discuss the causes, characteristics, and occurrence rules of the fake crackle and establish a reference for improving the reliability of the electronic stethoscope in lung auscultation. Methods: A total of 56 participants with healthy lungs (no underlying pulmonary disease, no recent respiratory symptoms, and no adventitious lung sound, as confirmed by an acoustic stethoscope) were enrolled in this study. A 30-s audio recording was recorded from each of the nine locations of the larynx and lungs of each participant with a 3M Littmann 3200 electronic stethoscope, and the audio was output in diaphragm mode and auscultated by the clinician. The doctor identified the fake crackles and analyzed their frequency spectrum. High-pass and low-pass filters were used to detect the frequency distribution of the fake crackles. Finally, the fake crackle was artificially regenerated to explore its causes. Results: A total of 500 audio recordings were included in the study, with 61 fake crackle audio recordings. Fake crackles were found predominantly in the lower lung. There were significant differences between lower lung and larynx (p < 0.001), lower lung and upper lung (p = 0.005), lower lung and middle lung (p = 0.005), and lower lung and infrascapular region (p = 0.027). Furthermore, more than 90% of fake crackles appeared in the inspiratory phase, similar to fine crackles, significantly interfering with clinical diagnosis. The spectral analysis revealed that the frequency range of fake crackles was approximately 250-1950 Hz. The fake crackle was generated when the diaphragm of the electronic stethoscope left the skin slightly but not completely. Conclusion: Fake crackles are most likely to be heard when using an electronic stethoscope to auscultate bilateral lower lungs, and the frequency of a fake crackle is close to that of a crackle, likely affecting the clinician's diagnosis.

SPRSound: Open-Source SJTU Paediatric Respiratory Sound Database

It has proved that the auscultation of respiratory sound has advantage in early respiratory diagnosis. Various methods have been raised to perform automatic respiratory sound analysis to reduce subjective diagnosis and physicians' workload. However, these methods highly rely on the quality of respiratory sound database. In this work, we have developed the first open-access paediatric respiratory sound database, SPRSound. The database consists of 2,683 records and 9,089 respiratory sound events from 292 participants. Accurate label is important to achieve a good prediction for adventitious respiratory sound classification problem. A custom-made sound label annotation software (SoundAnn) has been developed to perform sound editing, sound annotation, and quality assurance evaluation. A team of 11 experienced paediatric physicians is involved in the entire process to establish golden standard reference for the dataset. To verify the robustness and accuracy of the classification model, we have investigated the effects of different feature extraction methods and machine learning classifiers on the classification performance of our dataset. As such, we have achieved a score of 75.22%, 61.57%, 56.71%, and 37.84% for the four different classification challenges at the event level and record level.

Abnormal Respiratory Sounds Classification Using Deep CNN Through Artificial Noise Addition

Respiratory sound (RS) attributes and their analyses structure a fundamental piece of pneumonic pathology, and it gives symptomatic data regarding a patient's lung. A couple of decades back, doctors depended on their hearing to distinguish symptomatic signs in lung audios by utilizing the typical stethoscope, which is usually considered a cheap and secure method for examining the patients. Lung disease is the third most ordinary cause of death worldwide, so; it is essential to classify the RS abnormality accurately to overcome the death rate. In this research, we have applied Fourier analysis for the visual inspection of abnormal respiratory sounds. Spectrum analysis was done through Artificial Noise Addition (ANA) in conjunction with different deep convolutional neural networks (CNN) to classify the seven abnormal respiratory sounds-both continuous (CAS) and discontinuous (DAS). The proposed framework contains an adaptive mechanism of adding a similar type of noise to unhealthy respiratory sounds. ANA makes sound features enough reach to be identified more accurately than the respiratory sounds without ANA. The obtained results using the proposed framework are superior to previous techniques since we simultaneously considered the seven different abnormal respiratory sound classes.

Fine crackles on chest auscultation in the early diagnosis of idiopathic pulmonary fibrosis: a prospective cohort study

Introduction

Idiopathic pulmonary fibrosis (IPF) is an interstitial lung disease (ILD) with a poor prognosis. Early diagnosis and treatment of IPF may increase lifespan and preserve quality of life. Chest CT is the best test to diagnose IPF, but it is expensive and impractical as a screening test. Fine crackles on chest auscultation may be the only best to screen for IPF.

Methods

We prospectively assessed the presence and type of crackles on chest auscultation in all patients referred to the ILD Clinic at the Kingston Health Sciences Center in Ontario, Canada. Clinicians with varying levels of experience recorded the presence of fine crackles, coarse crackles or both independently and unaware of the final diagnosis. We applied multinomial logistic regression to adjust for ILD severity and factors that could affect the identification of crackles.

Results

We evaluated 290 patients referred to the ILD Clinic. On initial presentation, 93% of patients with IPF and 73% of patients with non-IPF ILD had fine crackles on auscultation. In patients with IPF, fine crackles were more common than cough (86%), dyspnoea (80%), low diffusing capacity (87%), total lung capacity (57%) and forced vital capacity (50%). There was 90% observer agreement in identifying fine crackles at a subsequent visit. In multiple regression analysis, the identification of fine crackles was unaffected by lung function, symptoms, emphysema, chronic obstructive pulmonary disease, obesity or clinician experience (p>0.05).

Conclusions

Fine crackles on chest auscultation are a sensitive and robust screening tool that can lead to early diagnosis and treatment of patients with IPF.

Unwrapping the phase portrait features of adventitious crackle for auscultation and classification: a machine learning approach

The paper delves into the plausibility of applying fractal, spectral, and nonlinear time series analyses for lung auscultation. The thirty-five sound signals of bronchial (BB) and pulmonary crackle (PC) analysed by fast Fourier transform and wavelet not only give the details of number, nature, and time of occurrence of the frequency components but also throw light onto the embedded air flow during breathing. Fractal dimension, phase portrait, and sample entropy help in divulging the greater randomness, antipersistent nature, and complexity of airflow dynamics in BB than PC. The potential of principal component analysis through the spectral feature extraction categorises BB, fine crackles, and coarse crackles. The phase portrait feature-based supervised classification proves to be better compared to the unsupervised machine learning technique. The present work elucidates phase portrait features as a better choice of classification, as it takes into consideration the temporal correlation between the data points of the time series signal, and thereby suggesting a novel surrogate method for the diagnosis in pulmonology. The study suggests the possible application of the techniques in the auscultation of coronavirus disease 2019 seriously affecting the respiratory system.

A machine-learning based approach to quantify fine crackles in the diagnosis of interstitial pneumonia: A proof-of-concept study

Fine crackles are frequently heard in patients with interstitial lung diseases (ILDs) and are known as the sensitive indicator for ILDs, although the objective method for analyzing respiratory sounds including fine crackles is not clinically available. We have previously developed a machine-learning-based algorithm which can promptly analyze and quantify the respiratory sounds including fine crackles. In the present proof-of-concept study, we assessed the usefulness of fine crackles quantified by this algorithm in the diagnosis of ILDs.We evaluated the fine crackles quantitative values (FCQVs) in 60 participants who underwent high-resolution computed tomography (HRCT) and chest X-ray in our hospital. Right and left lung fields were evaluated separately.In sixty-seven lung fields with ILDs in HRCT, the mean FCQVs (0.121 ± 0.090) were significantly higher than those in the lung fields without ILDs (0.032 ± 0.023, P < .001). Among those with ILDs in HRCT, the mean FCQVs were significantly higher in those with idiopathic pulmonary fibrosis than in those with other types of ILDs (P = .002). In addition, the increased mean FCQV was associated with the presence of traction bronchiectasis (P = .003) and honeycombing (P = .004) in HRCT. Furthermore, in discriminating ILDs in HRCT, an FCQV-based determination of the presence or absence of fine crackles indicated a higher sensitivity compared to a chest X-ray-based determination of the presence or absence of ILDs.We herein report that the machine-learning-based quantification of fine crackles can predict the HRCT findings of lung fibrosis and can support the prompt and sensitive diagnosis of ILDs.

A dataset for systematic testing of crackle separation techniques

Pulmonary crackles are indicative of lung pathology and may be used for diagnosis and monitoring of disease. Many algorithms have been proposed to separate the crackle sounds from the breath noise, but a lack of standardized processes for evaluating their performance makes comparisons difficult. In this paper we propose a standard data set to be used for systematic comparative testing.

Influence of observer preferences and auscultatory skill on the choice of terms to describe lung sounds: a survey of staff physicians, residents and medical students

BACKGROUND

In contrast with the technical progress of the stethoscope, lung sound terminology has remained confused, weakening the usefulness of auscultation. We examined how observer preferences regarding terminology and auscultatory skill influenced the choice of terms used to describe lung sounds.

METHODS

Thirty-one staff physicians (SP), 65 residents (R) and 47 medical students (MS) spontaneously described the audio recordings of 5 lung sounds classified acoustically as: (1) normal breath sound; (2) wheezes; (3) crackles; (4) stridor and (5) pleural friction rub. A rating was considered correct if a correct term or synonym was used to describe it (term use ascribed to preference). The use of any incorrect terms was ascribed to deficient auscultatory skill.

RESULTS

Rates of correct sound identification were: (i) normal breath sound: SP=21.4%; R=11.6%; MS=17.1%; (ii) wheezes: SP=82.8%; R=85.2%; MS=86.4%; (iii) crackles: SP=63%; R=68.5%; MS=70.7%; (iv) stridor: SP=92.8%; R=90%; MS=72.1% and (v) pleural friction rub: SP=35.7%; R=6.2%; MS=3.2%. The 3 groups used 66 descriptive terms: 17 were ascribed to preferences regarding terminology, and 49 to deficient auscultatory skill. Three-group agreement on use of a term occurred on 107 occasions: 70 involved correct terms (65.4%) and 37 (34.6%) incorrect ones. Rate of use of recommended terms, rather than accepted synonyms, was 100% for the wheezes and the stridor, 55% for the normal breath sound, 22% for the crackles and 14% for the pleural friction rub.

CONCLUSIONS

The observers' ability to describe lung sounds was high for the wheezes and the stridor, fair for the crackles and poor for the normal breath sound and the pleural friction rub. Lack of auscultatory skill largely surpassed observer preference as a factor determining the choice of terminology. Wide dissemination of educational programs on lung auscultation (eg, self-learning via computer-assisted learning tools) is urgently needed to promote use of standardised lung sound terminology.

The acoustic characteristics of fine crackles predict honeycombing on high-resolution computed tomography

Background

Honeycombing on high-resolution computed tomography (HRCT) is a distinguishing feature of usual interstitial pneumonia and predictive of poor outcome in interstitial lung diseases (ILDs). Although fine crackles are common in ILD patients, the relationship between their acoustic features and honeycombing on HRCT has not been well characterized.

Methods

Lung sounds were digitally recorded from 71 patients with fine crackles and ILD findings on chest HRCT. Lung sounds were analyzed by fast Fourier analysis using a sound spectrometer (Easy-LSA; Fukuoka, Japan). The relationships between the acoustic features of fine crackles in inspiration phases (onset timing, number, frequency parameters, and time-expanded waveform parameters) and honeycombing in HRCT were investigated using multivariate logistic regression analysis.

Results

On analysis, the presence of honeycombing on HRCT was independently associated with onset timing (early vs. not early period; odds ratios [OR] 10.407, 95% confidence interval [95% CI] 1.366–79.298, P = 0.024), F99 value (the percentile frequency below which 99% of the total signal power is accumulated) (unit Hz = 100; OR 5.953, 95% CI 1.221–28.317, P = 0.029), and number of fine crackles in the inspiratory phase (unit number = 5; OR 4.256, 95% CI 1.098–16.507, P = 0.036). In the receiver-operating characteristic curves for number of crackles and F99 value, the cutoff levels for predicting the presence of honeycombing on HRCT were calculated as 13.2 (area under the curve [AUC], 0.913; sensitivity, 95.8%; specificity, 75.6%) and 752 Hz (AUC, 0.911; sensitivity, 91.7%; specificity, 85.2%), respectively. The multivariate logistic regression analysis additionally using these cutoff values revealed an independent association of number of fine crackles in the inspiratory phase, F99 value, and onset timing with the presence of honeycombing (OR 33.907, 95% CI 2.576–446.337, P = 0.007; OR 19.397, 95% CI 2.311–162.813, P = 0.006; and OR 12.383, 95% CI 1.443–106.293, P = 0.022; respectively).

Conclusions

The acoustic properties of fine crackles distinguish the honeycombing from the non-honeycombing group. Furthermore, onset timing, number of crackles in the inspiratory phase, and F99 value of fine crackles were independently associated with the presence of honeycombing on HRCT. Thus, auscultation routinely performed in clinical settings combined with a respiratory sound analysis may be predictive of the presence of honeycombing on HRCT.

Detection of inspiratory recruitment of atelectasis by automated lung sound analysis as compared to four-dimensional computed tomography in a porcine lung injury model

Background

Cyclic recruitment and de-recruitment of atelectasis (c-R/D) is a contributor to ventilator-induced lung injury (VILI). Bedside detection of this dynamic process could improve ventilator management. This study investigated the potential of automated lung sound analysis to detect c-R/D as compared to four-dimensional computed tomography (4DCT).

Methods

In ten piglets (25 ± 2 kg), acoustic measurements from 34 thoracic piezoelectric sensors (Meditron ASA, Norway) were performed, time synchronized to 4DCT scans, at positive end-expiratory pressures of 0, 5, 10, and 15 cmH₂O during mechanical ventilation, before and after induction of c-R/D by surfactant washout. 4DCT was post-processed for within-breath variation in atelectatic volume (Δ atelectasis) as a measure of c-R/D. Sound waveforms were evaluated for: 1) dynamic crackle energy (dCE): filtered crackle sounds (600–700 Hz); 2) fast Fourier transform area (FFT area): spectral content above 500 Hz in frequency and above −70 dB in amplitude in proportion to the total amount of sound above −70 dB amplitude; and 3) dynamic spectral coherence (dSC): variation in acoustical homogeneity over time. Parameters were analyzed for global, nondependent, central, and dependent lung areas.

Results

In healthy lungs, negligible values of Δ atelectasis, dCE, and FFT area occurred. In lavage lung injury, the novel dCE parameter showed the best correlation to Δ atelectasis in dependent lung areas (R² = 0.88) where c-R/D took place. dCE was superior to FFT area analysis for each lung region examined. The analysis of dSC could predict the lung regions where c-R/D originated.

Conclusions

c-R/D is associated with the occurrence of fine crackle sounds as demonstrated by dCE analysis. Standardized computer-assisted analysis of dCE and dSC seems to be a promising method for depicting c-R/D.

Electronic supplementary material

The online version of this article (10.1186/s13054-018-1964-6) contains supplementary material, which is available to authorized users.

Automatic adventitious respiratory sound analysis: A systematic review

Background

Automatic detection or classification of adventitious sounds is useful to assist physicians in diagnosing or monitoring diseases such as asthma, Chronic Obstructive Pulmonary Disease (COPD), and pneumonia. While computerised respiratory sound analysis, specifically for the detection or classification of adventitious sounds, has recently been the focus of an increasing number of studies, a standardised approach and comparison has not been well established.

Objective

To provide a review of existing algorithms for the detection or classification of adventitious respiratory sounds. This systematic review provides a complete summary of methods used in the literature to give a baseline for future works.

Data sources

A systematic review of English articles published between 1938 and 2016, searched using the Scopus (1938-2016) and IEEExplore (1984-2016) databases. Additional articles were further obtained by references listed in the articles found. Search terms included adventitious sound detection, adventitious sound classification, abnormal respiratory sound detection, abnormal respiratory sound classification, wheeze detection, wheeze classification, crackle detection, crackle classification, rhonchi detection, rhonchi classification, stridor detection, stridor classification, pleural rub detection, pleural rub classification, squawk detection, and squawk classification.

Study selection

Only articles were included that focused on adventitious sound detection or classification, based on respiratory sounds, with performance reported and sufficient information provided to be approximately repeated.

Data extraction

Investigators extracted data about the adventitious sound type analysed, approach and level of analysis, instrumentation or data source, location of sensor, amount of data obtained, data management, features, methods, and performance achieved.

Data synthesis

A total of 77 reports from the literature were included in this review. 55 (71.43%) of the studies focused on wheeze, 40 (51.95%) on crackle, 9 (11.69%) on stridor, 9 (11.69%) on rhonchi, and 18 (23.38%) on other sounds such as pleural rub, squawk, as well as the pathology. Instrumentation used to collect data included microphones, stethoscopes, and accelerometers. Several references obtained data from online repositories or book audio CD companions. Detection or classification methods used varied from empirically determined thresholds to more complex machine learning techniques. Performance reported in the surveyed works were converted to accuracy measures for data synthesis.

Limitations

Direct comparison of the performance of surveyed works cannot be performed as the input data used by each was different. A standard validation method has not been established, resulting in different works using different methods and performance measure definitions.

Conclusion

A review of the literature was performed to summarise different analysis approaches, features, and methods used for the analysis. The performance of recent studies showed a high agreement with conventional non-automatic identification. This suggests that automated adventitious sound detection or classification is a promising solution to overcome the limitations of conventional auscultation and to assist in the monitoring of relevant diseases.

Auscultation of the respiratory system

Auscultation of the lung is an important part of the respiratory examination and is helpful in diagnosing various respiratory disorders. Auscultation assesses airflow through the trachea-bronchial tree. It is important to distinguish normal respiratory sounds from abnormal ones for example crackles, wheezes, and pleural rub in order to make correct diagnosis. It is necessary to understand the underlying pathophysiology of various lung sounds generation for better understanding of disease processes. Bedside teaching should be strengthened in order to avoid erosion in this age old procedure in the era of technological explosion.

Computerized respiratory sounds in patients with COPD: a systematic review

Computerized respiratory sound analysis provides objective information about the respiratory system and may be useful to monitor patients with chronic obstructive pulmonary disease (COPD) and detect exacerbations early. For these purposes, a thorough understanding of the typical computerized respiratory sounds in patients with COPD during stable periods is essential. This review aimed to systematize the existing evidence on computerized respiratory sounds in stable COPD. A literature search in the Medline, EBSCO, Web of Knowledge and Scopus databases was performed. Seven original articles were included. The maximum frequencies of normal inspiratory sounds at the posterior chest were between 113 and 130Hz, lower than the frequency found at trachea (228 Hz). During inspiration, the frequency of normal respiratory sounds was found to be higher than expiration (130 vs. 100Hz). Crackles were predominantly inspiratory (2.9-5 vs. expiratory 0.73-2) and characterized by long durations of the variables initial deflection width (1.88-2.1 ms) and two cycle duration (7.7-11.6 ms). Expiratory wheeze rate was higher than inspiratory rate. In patients with COPD normal respiratory sounds seem to follow the pattern observed in healthy people and adventitious respiratory sounds are mainly characterized by inspiratory and coarse crackles and expiratory wheezes. Further research with larger samples and following the Computerized Respiratory Sound Analysis (CORSA) guidelines are needed.

Modeling of pulmonary crackles using wavelet networks

In this study, wavelet networks are used to model pulmonary crackles with a view to extract features for the classification analysis of crackles obtained from subjects with a wide spectrum of pulmonary disorders. Crackles are very common adventitious sounds which are transient in character and whose characteristics, such as type, number of occurrence and pitch, convey information regarding the type and severity of the pulmonary disease. Crackles generally start with a sharp deflection and continue with a damped and progressively wider sinusoidal wave. In this study, due to the capability of time-frequency representation of wavelet functions, wavelet network (WN) is employed to characterize crackles, and the parameters acquired from wavelet nodes are used to distinguish them into two clinical classes, i.e. fine and coarse. For this purpose, a wavelet function (complex Morlet) in the first node is trained to fit the crackles and the second wavelet node is tuned to represent the error of the first node. Both of the nodes are, then, trained to minimize the total representative error. The five parameters of the WN node, i.e. scaling, time-shifting, frequency and two weight factors of sinus and cosines components are used as features in the classification analysis of crackles. The crackle information is strongly represented by the first wavelet node, therefore, the parameters belonging to the first node are used in the classification of crackles.

Early detection of deteriorating ventilation by monitoring bilateral chest wall dynamics in the rabbit

PURPOSE

Mechanical complications during assisted ventilation can evolve due to worsening lung disease or problems in airway management. These complications affect lung compliance or airway resistance, which in turn affect the chest wall dynamics. The objective of this study was to explore the utility of continuous monitoring of the symmetry and dynamics of chest wall motion in the early detection of complications during mechanical ventilation.

METHODS

The local tidal displacement (TDi) values of each side of the chest and epigastrium were measured by three miniature motion sensors in 18 rabbits. The TDi responses to changes in peak inspiratory pressure (n = 7), induction of one-lung intubation (n = 7), and slowly progressing pneumothorax (PTX) (n = 6) were monitored in parallel with conventional respiratory (SpO(2), EtCO(2), pressure and flow) and hemodynamic (HR and BP) indices. PTX was induced by injecting air into the pleural space at a rate of 1 mL/min.

RESULTS

A strong correlation (R(2) = 0.99) with a slope close to unity (0.94) was observed between percent change in tidal volume and in TDi. One-lung ventilation was identified by conspicuous asymmetry development between left and right TDis. These indices provided significantly early detection of uneven ventilation during slowly developing PTX (within 12.9 ± 6.6 min of onset, p = 0.02) almost 1 h before the SpO(2) dropped (77.3 ± 27.4 min, p = 0.02). Decreases in TDi of the affected side paralleled the progression of PTX.

CONCLUSIONS

Monitoring the local TDi is a sensitive method for detecting changes in tidal volume and enables early detection of developing asymmetric ventilation.

CLASSIFICATION OF RESPIRATORY SOUNDS BY USING AN ARTIFICIAL NEURAL NETWORK

In this paper, a classification method for respiratory sounds (RSs) in patients with asthma and in healthy subjects is presented.

Wavelet transform is applied to a window containing 256 samples. Elements of the feature vectors are obtained from the wavelet coefficients. The best feature elements are selected by using dynamic programming. Grow and Learn (GAL) neural network, Kohonen network and multi-layer perceptron (MLP) are used for the classification.

It is observed that RSs of patients (with asthma) and healthy subjects are successfully classified by the GAL network.

Probing the existence of medium pulmonary crackles via model-based clustering

The objective of this study is to probe the existence of a third crackle type, medium, besides the traditionally accepted types, namely, fine and coarse crackles and, furthermore, to explore the representative parameter values for each crackle type. A set of clustering experiments have been conducted on pulmonary crackles to this end. A model-based clustering algorithm, the Expectation-Maximization algorithm, is used and the resulting cluster numbers are validated with Bayesian Inference Criterion. Four different feature sets are extracted from the preprocessed crackle samples, the first of which consists of conventional parameters derived from the zero-crossings of crackle waveforms. The second feature set corresponds to the spectral components of the crackles, whereas the remaining two sets are derived from a single- and double-nodes wavelet network modeling. The results of the clustering experiments demonstrate strong evidence for the existence of a third crackle type. Moreover the labels yielded by clustering experiments, using different feature sets match for roughly 80% or more of the crackle samples, resulting in similar representative crackle parameter values of the three clusters for all feature sets.

Evaluation of the usefulness of spectral analysis of inspiratory lung sounds recorded with phonopneumography in patients with interstitial pneumonia

PURPOSE

We investigated whether spectral analysis with fast Fourier transform (FFT) of inspiratory lung sounds is useful in the diagnosis and evaluation of the severity of interstitial pneumonia (IP).

SUBJECTS AND METHODS

The study population included 10 healthy volunteers (healthy group) and 21 patients with IP (IP group). We generated inspiratory averaged linear intensities using FFT and determined frequency at maximum sound intensity (Fmax), and quartile frequencies (f25, f50, and f75), compared these values between the groups, generated receiver operating characteristic curves to compare the detectability of IP between the indices and auscultation in all cases, and tested for the correlation of these indices with pulmonary function tests and the fibrosis scores from high-resolution computed tomography images assessed by 3 observers.

RESULTS

Both f50 and f75 were significantly higher in the IP group, but their ability to detect IP was inferior to that of auscultation. They had negative correlations with percent vital capacity and had positive correlations with the fibrosis scores calculated by the 3 different observers.

DISCUSSION

These results were considered to reflect the presence of fine crackles and alterations in pulmonary sound-conduction characteristics caused by IP and indicate that spectral analysis of lung sounds is useful in the diagnosis and evaluation of the severity of IP, although future study is necessary to improve its utility.

Analysis of discontinuous adventitious lung sounds by Hilbert-Huang spectrum

It is now widely accepted that crackles are associated with different pulmonary pathologies and different efforts have been done to detect and to extract them. Consequently, due to the difficulty for their characterization, the selection of an adequate time-frequency representation (TFR) for the analysis of their time-frequency dynamics is important. Traditionally, normal and abnormal lung sounds have been analyzed by the Spectrogram (SP). However, this analysis tool has certain disadvantages when one deals with nonstationary signals. As an effort to point out the appropriate analysis tool for crackles, this paper shows the performance of the Hilbert-Huang spectrum (HHS) for the analysis of fine and coarse crackles, simulated and real ones. The HHS allowed to analyze the evolving time-frequency of crackle sounds straightforward with good resolution compared with SP. Beside this enhanced time-frequency course, HHS could be useful to establish a signature to detect and separate fine from coarse crackles in order to screen pathologies and their progress during medication.

A texture-based classification of crackles and squawks using lacunarity

An automatic classification method to efficiently discriminate the types of discontinuous breath sounds (DBSs), i.e., fine crackles (FCs), coarse crackles (CC), and squawks (SQ), is presented in this paper. Using the lacunarity of the acquired DBS, the proposed classification method, namely LAC, introduces a texture-based approach that captures the differences in the distribution of FC, CC, and SQ across the breathing cycle, which may lead to more accurate characterization of the pulmonary acoustical changes due to the related pathology. Prior to the lacunarity analysis, wavelet-based denoising of DBS is employed to eliminate effects of the vesicular sound (background noise) to DBS oscillatory pattern. LAC analysis builds its classification power both upon the use of lacunarity at an optimum scale and the approximation of its trajectory across an optimum range of scales using a three-parameter hyperbola model. LAC is applied to 363 DBS corresponding to 25 cases included in four lung sound databases. Results show that LAC efficiently classifies the three DBS categories in the comparison groups of FC-CC, FC-SQ (both with mean accuracy of 100%), CC-SQ (mean accuracy of 99.62%-100%), and FC-CC-SQ (mean accuracy of 99.75%-100%). When compared to other classification tools, LAC seems quite attractive, since, without employing high computational complexity, it results in high classification accuracy. Moreover, LAC introduces a "texture" concept in the analysis of breath sounds, something that strongly relates to the perception of the bioacoustic signals by the physician. Due to its simplicity, LAC could be implemented in a real-time context and be used in clinical medicine as a module of an integrated intelligent patient evaluation system.

Crackles detection using a time-variant autoregressive model

Several techniques have been explored to detect automatically fine and coarse crackles; however, the solution for automatic detection of crackles remains insufficient. The purpose of this work was to explore the capacity of the time-variant autoregressive (TVAR) model to detect and to provide an estimate number of fine and coarse crackles in lung sounds. Thus, simulated crackles inserted in normal lung sounds and real lung sounds containing adventitious sounds were processed with TVAR and by an expert that based crackle detection on time-expanded waveform-analysis. The coefficients of the TVAR were obtained by an adaptive filtering prediction scheme. The adaptive filter used the recursive least squares algorithm with a forgetting factor of 0.97 and the model order was four. TVAR model showed an efficiency to detect crackles over 90% even with crackles overlapping and amplitudes as low as 1.5 of the standard deviation of background lung sounds, where expert presented an efficiency around 30%. In conclusion, TVAR model is a proper alternative to detect and to provide an estimate number of fine and coarse crackles, even in presence of crackles overlapping and crackles with low amplitude, conditions where crackles detection based on time-expanded waveform-analysis reveals evident limitations.

Design, construction, and evaluation of a bioacoustic transducer testing (BATT) system for respiratory sounds

Many different transducers are employed for recording respiratory sounds including accelerometers and microphones in couplers. However, there is no standard lung sound transducer or any device to compare transducers so that measurements from different laboratories can be determined to be of physiologic origin rather than technical artifacts of the transducers. To address this problem, we designed and constructed a prototype of a device that can be used to compare accelerometers, microphones enclosed in couplers, and stethoscopes. The prototype device consists of a rigid chamber containing a loudspeaker that opens to an antechamber covered by a viscoelastic material with mechanical properties similar to human skin and subcutaneous tissue. When driven by a white noise source, we found the sound output at the surface to be useful to comparatively evaluate sensors between 100 and 1200 Hz where lung sounds have most of their spectral energy. We compared the viscoelastic layer to similar thicknesses of fresh meat and fat and found them to produce similar acoustic spectra. This device allows air-coupled transducers, accelerometers, and stethoscopes used in respiratory sounds measurements to be compared under physical conditions similar to their intended use.

Comparison of lung sound transducers using a bioacoustic transducer testing system

Sensors used for lung sound research are generally designed by the investigators or adapted from devices used in related fields. Their relative characteristics have never been defined. We employed an artificial chest wall with a viscoelastic surface and a white noise signal generator as a stable source of sound to compare the frequency response and pulse waveform reproduction of a selection of devices used for lung sound research. We used spectral estimation techniques to determine frequency response and cross-correlation of pulses to determine pulse shape fidelity. The sensors evaluated were the Siemens EMT 25 C accelerometer (Siemens); PPG 201 accelerometer (PPG); Sony ECM-T150 electret condenser microphone with air coupler (air coupler; with cylindrical air chambers of 5-, 10-, and 15-mm diameter and conical air chamber of 10-mm diameter); Littman classic stethoscope head (Littman) connected to an electret condenser microphone; and the Andries Tek (Andries) electronic stethoscope. We found that the size and shape of the air coupler chamber to have no important effect on the detected sound. The Siemens, air coupler, and Littman performed similarly with relatively flat frequency responses from 200 to 1,200 Hz. The PPG had the broadest frequency response, with useful sensitivity extending to 4,000 Hz. The Andries' frequency response was the poorest above 1,000 Hz. Accuracy in reproducing pulses roughly corresponded with the high-frequency sensitivity of the sensors. We conclude that there are important differences among commonly used lung sound sensors that have to be defined to allow the comparison of data from different laboratories.

Discrimination analysis of discontinuous breath sounds using higher-order crossings

The paper evaluates the performance of an automatic discrimination analysis (DA) method used to discriminate efficiently the types of discontinuous breath sound (DBS), i.e. fine crackles (FCs), coarse crackles (CCs) and squawks (SQs); this may lead to more accurate characterisation of the pulmonary acoustical changes due to the related pathology. Based on higher-order crossings (HOCs), the proposed method, HOC-DA, captured the differences in the oscillatory patterns of FCs, CCs and SQs, which are only exposed when higher (> 1) crossings are employed. Prior to HOC-DA, wavelet-based de-noising of DBSs was employed to eliminate the effects of the vesicular sound (background noise) from their oscillatory pattern. The HOC-DA was applied to 157 discontinuous breath sounds corresponding to 16 cases included in three lung sound databases. Results showed that the HOC-DA efficiently separated FCs from CCs, SQs from CCs (both with an accuracy of 100%), and SQs from FCs (accuracy of 80%), with the optimum order ranging from 9 to 11. When compared with other classification tools, the HOC-DA resulted in high discrimination accuracy without involving high computational complexity. Owing to its simplicity, it could be implemented in a real-time context and be used in clinical medicine as a module of an integrated intelligent patient evaluation system.

Auditory detection of simulated crackles in breath sounds

BACKGROUND

Computerized analysis of breath sounds has relied on human auditory perception as the reference standard for identifying crackles. In this study, we tested the human audibility of crackles by superimposing artificial clicks on recorded breath sounds and having physicians listen to the recordings to see if they could identify the crackles.

OBJECTIVES

To establish the audibility of simulated crackles introduced in breath sounds of different intensity, to study the effects of crackle characteristics on their audibility, and to investigate crackle detection within and between observers.

METHODS

Fine, medium, and coarse crackles with large and small amplitude were synthesized by computer software. Waveform parameters were based on published characteristics of lung sound crackles. The amplitude for small crackles was defined as just above the threshold of audibility for simulated crackles inserted in sound recorded during breath hold. Simulated crackles were then superimposed on breath sounds recorded at 0 L/s (breath hold), 1 L/s, and 2 L/s airflow. Five physicians listened during playback on two separate occasions to determine if crackles could be heard and to calculate the interobserver and intraobserver variations.

RESULTS

Failed detection of crackles was significantly more common in the following conditions: (1) background breath sounds had higher intensity (2 L/s airflow) compared to lower intensity (1 L/s), (2) crackle type was coarse or medium compared to fine, and (3) crackle amplitude was small compared to large. Both intraobserver and interobserver agreements were high (kappa > 0.6).

RELEVANCE

The validation of automated techniques for crackle detection in lung sound analysis should not rely on auscultation as the only reference. Detection of crackles is facilitated when patients take slow, deep breaths that generate little breath sounds.

The influence of "sputum-like gel" viscosity on crackle characteristics in a mechanically ventilated porcine lung model

This study investigated the influence of airway secretion viscosity on the characteristics of crackle sounds produced using a mechanically ventilated porcine lung model. Aqueous ultrasonic methylene blue stained gel solutions of viscosity 100, 150 and 200 P were prepared and instilled into 15 isolated, mechanically ventilated, porcine lungs immersed in water. Sound signals recorded by a hydrophone before and after instillation of gel were subjected to both fast Fourier transform and wave-form analysis. At the completion of sound recording, the main bronchi were dissected and the location of the stained gel was photographically recorded. Wave-form analysis demonstrated that expiratory phase crackle incidence and amplitude were both significantly higher than inspiratory phase data. This study demonstrates that crackle duration and amplitude are inversely related to gel viscosity and that electronic lung sound analysis can provide indirect evidence of sputum viscosity.

Representation and classification of breath sounds recorded in an intensive care setting using neural networks

OBJECTIVE

Develop and test methods for representing and classifying breath sounds in an intensive care setting.

METHODS

Breath sounds were recorded over the bronchial regions of the chest. The breath sounds were represented by their averaged power spectral density, summed into feature vectors across the frequency spectrum from 0 to 800 Hertz. The sounds were segmented by individual breath and each breath was divided into inspiratory and expiratory segments. Sounds were classified as normal or abnormal. Different back-propagation neural network configurations were evaluated. The number of input features, hidden units, and hidden layers were varied.

RESULTS

2127 individual breath sounds from the ICU patients and 321 breaths from training tapes were obtained. Best overall classification rate for the ICU breath sounds was 73% with 62% sensitivity and 85% specificity. Best overall classification rate for the training tapes was 91% with 87% sensitivity and 95% specificity.

CONCLUSIONS

Long term monitoring of lung sounds is not feasible unless several barriers can be overcome. Several choices in signal representation and neural network design greatly improved the classification rates of breath sounds. The analysis of transmitted sounds from the trachea to the lung is suggested as an area for future study.

Auscultation revisited: the waveform and spectral characteristics of breath sounds during general anesthesia

Although auscultation is commonly used as a continuous monitoring tool during anesthesia, the breath sounds of anesthetized patients have never been systematically studied. In this investigation we used digital audio technology to record and analyze the breath sounds of 14 healthy adult patients receiving general anesthesia with positive pressure ventilation. Sounds recorded from inside the esophagus were compared to those recorded from the surface of the chest, and corresponding airflow was measured with a pneumotachograph. The sound samples associated with inspiratory and expiratory phases were analyzed in the time domain (RMS amplitude) and frequency domain (peak frequency, spectral edge, and power ratios). There was a positive linear correlation (R2 > 0.9) between inspiratory flow and sound amplitude in the precordial and esophageal samples of all patients. The RMS amplitude of the inspiratory and expiratory sounds was approximately 13 times greater when recorded from inside the esophagus than from the surface of the chest in all patients at all flows (p < 0.001). The peak frequency (Hz) was significantly higher in the esophageal recordings than the precordial samples (298 +/- 9 vs 181 +/- 10, P < 0.0001), as was the 97% spectral edge (Hz) (740 +/- 7 vs 348 +/- 16, P < 0.0001). In the adult population esophageal stethoscopes yield higher frequencies and greater amplitude than precordial stethoscopes. Quantification of lung sounds may provide for improved monitoring and diagnostic capability during anesthesia and surgery.

Computer-based lung sound simulation

An algorithm for the simulation of normal and pathological lung sounds is developed. The simulation algorithm is implemented on a personal computer as well as on a digital signal processor system in real time. Normal, bronchial and tracheal breathing sounds can be generated, and continuous and discontinuous adventitious lung sounds can be added. The attributes of the individual sound components, such as loudness, frequency, duration or number of occurrences within one breathing cycle, are controlled by the user. The quality of the simulations is evaluated by sending audio tapes to 15 experienced pulmonary physicians for a formal assessment. Each tape contains five simulated lung sounds and five real lung sounds from a commercially available teaching tape, presented in random order. Simulated lung sounds are slightly better rated in terms of realism and signal quality when compared to the recordings from the teaching tape. The differences are, however, not significant. 13 out of the 15 physicians feel that computer-based lung sound simulators would be a useful and desirable teaching tool for auscultation courses.