Adventitious Respiratory Sounds Research Articles

Auscultation-Based Pulmonary Disease Detection through Parallel Transformation and Deep Learning.

Respiratory diseases are among the leading causes of death, with many individuals in a population frequently affected by various types of pulmonary disorders. Early diagnosis and patient monitoring (traditionally involving lung auscultation) are essential for the effective management of respiratory diseases. However, the interpretation of lung sounds is a subjective and labor-intensive process that demands considerable medical expertise, and there is a good chance of misclassification. To address this problem, we propose a hybrid deep learning technique that incorporates signal processing techniques. Parallel transformation is applied to adventitious respiratory sounds, transforming lung sound signals into two distinct time-frequency scalograms: the continuous wavelet transform and the mel spectrogram. Furthermore, parallel convolutional autoencoders are employed to extract features from scalograms, and the resulting latent space features are fused into a hybrid feature pool. Finally, leveraging a long short-term memory model, a feature from the latent space is used as input for classifying various types of respiratory diseases. Our work is evaluated using the ICBHI-2017 lung sound dataset. The experimental findings indicate that our proposed method achieves promising predictive performance, with average values for accuracy, sensitivity, specificity, and F1-score of 94.16%, 89.56%, 99.10%, and 89.56%, respectively, for eight-class respiratory diseases; 79.61%, 78.55%, 92.49%, and 78.67%, respectively, for four-class diseases; and 85.61%, 83.44%, 83.44%, and 84.21%, respectively, for binary-class (normal vs. abnormal) lung sounds.

Classification of adventitious respiratory sounds through histogram of oriented gradients

Las enfermedades respiratorias representan una de las principales causas de muerte en el mundo. La prevención, el diagnóstico oportuno y el tratamiento efectivo son pilares fundamentales para reducir la propagación de enfermedades, así como su impacto negativo en la sociedad. El presente trabajo está enfocado en desarrollar un modelo computacional, a través del uso del algoritmo Histograma de Gradientes Orientados (HOG) en conjunto con técnicas de aprendizaje de máquina, capaz de clasificar sonidos respiratorios adventicios; para apoyar en el diagnóstico oportuno de enfermedades respiratorias. Para ello, se propone utilizar al HOG como extractor de características, el cual no ha sido explorado en la literatura actual sobre la clasificación de sonidos respiratorios. De igual manera, se plantea usar distintos algoritmos de aprendizaje de máquina, como: Máquina de Soporte Vectorial (SVM), K-Vecinos más Cercanos (KNN) y Bosques Aleatorios (RF). Asimismo, evaluar el modelo según el marco de trabajo proveído por la base de datos ICBHI17. La principal contribución del presente trabajo es el análisis de diferentes configuraciones del histograma de gradientes orientados para optimizar los modelos de aprendizaje de máquina. En donde, después de una serie de experimentos, nuestro mejor modelo surgió con la configuración Alternativa 4 (ALT 4); que obtuvo como resultados: 55.07% ACC, 34.37% BAL, 51.52% SCC, 75.87% SPE y 27.18% SEN.

Bayesian optimized GoogLeNet based respiratory signal prediction model from empirically decomposed gammatone visualization

Discrimination of normal and adventitious respiratory sounds from stethoscope auscultation by human ears is challenging owing to low frequency characteristics and varying frequency range for inspiration and expiration. This makes the diagnosis of pulmonary disorders a subjective one relying on the experience and hearing capability of the physician. Most computer assisted diagnosis systems formulated to address these limitations fail to capture the inherent acoustic property of respiratory sounds close to human auditory system. To circumvent this problem, this study exploits the gammatone filter banks to evaluate the distribution of all frequency components present in the signal. To categorize the respiratory signals, GoogLeNet based Convolutional Neural Networks (CNN) prediction model is developed through Time-Frequency (TF) visualization of gammatone cepstral coefficients acquired from decomposed Intrinsic Mode Functions (IMFs) in an empirical manner. As the performance of the CNN model is greatly dependent on the learning environment, this study also tends to optimize the values of the hyper parameters to enhance the classification performance of the CNN model. Accordingly, the optimal values of initial learning rate, L2 regularization and momentum are identified for both Stochastic Gradient Descent Momentum (SGDM) and Adaptive Momentum (ADAM) optimizer using Bayesian optimization technique. Experimentation is carried out for three TF visualization methods viz. spectrograms, scalograms and Constant Q spectrograms. Evaluated results show that scalogram method of classification yields high accuracy of 87.37% and 88.32% for default selection of hyperparameters using SGDM and ADAM optimizers and with optimal selection of hyperparameters based on objective function reveal an improved accuracy of 93.68 % and 95.67% for SGDM and ADAM optimizers.

Automatic wheeze segmentation using harmonic-percussive source separation and empirical mode decomposition.

Wheezes are adventitious respiratory sounds commonly present in patients with respiratory conditions. The presence of wheezes and their time location are relevant for clinical reasons, such as understanding the degree of bronchial obstruction. Conventional auscultation is usually employed to analyze wheezes, but remote monitoring has become a pressing need during recent years. Automatic respiratory sound analysis is required to reliably perform remote auscultation. In this work we propose a method for wheeze segmentation. Our method starts by decomposing a given audio excerpt into intrinsic mode frequencies using empirical mode decomposition. Then, we apply harmonic-percussive source separation to the resulting audio tracks and get harmonic-enhanced spectrograms, which are processed to obtain harmonic masks. Subsequently, a series of empirically derived rules are applied to find wheeze candidates. Finally, the candidates stemming from the different audio tracks are merged and median filtered. In the evaluation stage, we compare our method to three baselines on the ICBHI 2017 Respiratory Sound Database, a challenging dataset containing various noise sources and background sounds. Using the full dataset, our method outperforms the baselines, achieving an F1 of 41.9%. Our method's performance is also better than the baselines across several stratified results focusing on five variables: recording equipment, age, sex, body-mass index, and diagnosis. We conclude that, contrary to what has been reported in the literature, wheeze segmentation has not been solved for real life scenario applications. Adaptation of existing systems to demographic characteristics might be a promising step in the direction of algorithm personalization, which would make automatic wheeze segmentation methods clinically viable.

Reliability and Validity of Computerized Adventitious Respiratory Sounds in People with Bronchiectasis.

Background: Computerized adventitious respiratory sounds (ARS), such as crackles and wheezes, have been poorly explored in bronchiectasis, especially their measurement properties. This study aimed to test the reliability and validity of ARS in bronchiectasis. Methods: Respiratory sounds were recorded twice at 4 chest locations on 2 assessment sessions (7 days apart) in people with bronchiectasis and daily sputum expectoration. The total number of crackles, number of wheezes and wheeze occupation rate (%) were the parameters extracted. Results: 28 participants (9 men; 62 ± 12 y) were included. Total number of crackles and wheezes showed moderate within-day (ICC 0.87, 95% CI 0.74−0.94; ICC 0.86, 95% CI 0.71−0.93) and between-day reliability (ICC 0.70, 95% CI 0.43−0.86; ICC 0.78, 95% CI 0.56−0.90) considering all chest locations and both respiratory phases; wheeze occupation rate showed moderate within-day reliability (ICC 0.86, 95% CI 0.71−0.93), but poor between-day reliability (ICC 0.71, 95% CI 0.33−0.87). Bland−Altman plots revealed no systematic bias, but wide limits of agreement, particularly in the between-days analysis. All ARS parameters correlated moderately with the amount of daily sputum expectoration (r > 0.4; p < 0.05). No other significant correlations were observed. Conclusion: ARS presented moderate reliability and were correlated with the daily sputum expectoration in bronchiectasis. The use of sequential measurements may be an option to achieve greater accuracy when ARS are used to monitor or assess the effects of physiotherapy interventions in this population.

Lung auscultation: Pathogenic mechanisms underlying the respiratory sounds

Auscultation of the lungs is one of the oldest diagnostic methods invented by Rene Laenneck. Nowadays, auscultation hasn’t lost its clinical significance. However, the development of science and technical progress have brought a lot of new ideas about the pathogenesis and interpretation of lungs sounds. This, in turn, caused some changes in the nomenclature. Moreover, the nomenclature of lung sounds in several European countries does not differ significantly but does not fully correspond to the recommended terminology for the English language. In addition, the mechanisms underlying some respiratory murmurs are interpreted ambiguously at present. For example, the main respiratory sound is designated both as “vesicular” and as “normal” respiration because of persisting differences in explaining the pathogenesis of this murmur. There is also a considerable difference in the terminology of added sounds such as crackles, wheezing, rhonchi, as well as the difference in understanding of the underlying mechanisms.Aim. This review analyses the existing differences in the interpretation of lung auscultation and the terminology used to describe respiratory sounds in the Russian and foreign medical literature.Conclusion. Comparative analysis of Russian and foreign publications concerning the issues of lung auscultation demonstrates the existing differences in the terminology and understanding of the mechanisms that underly the main and adventitious respiratory sounds. Forming the common information space requires the unification of terminology and interpretation of physiological and pathological processes in the lungs responsible for the respiratory sounds. Recording the lung sounds with subsequent computer analysis will make it possible to objectify the auscultation data and classify the sounds more accurately.

Automatic Classification of Adventitious Respiratory Sounds: A (Un)Solved Problem?

(1) Background: Patients with respiratory conditions typically exhibit adventitious respiratory sounds (ARS), such as wheezes and crackles. ARS events have variable duration. In this work we studied the influence of event duration on automatic ARS classification, namely, how the creation of the Other class (negative class) affected the classifiers’ performance. (2) Methods: We conducted a set of experiments where we varied the durations of the other events on three tasks: crackle vs. wheeze vs. other (3 Class); crackle vs. other (2 Class Crackles); and wheeze vs. other (2 Class Wheezes). Four classifiers (linear discriminant analysis, support vector machines, boosted trees, and convolutional neural networks) were evaluated on those tasks using an open access respiratory sound database. (3) Results: While on the 3 Class task with fixed durations, the best classifier achieved an accuracy of 96.9%, the same classifier reached an accuracy of 81.8% on the more realistic 3 Class task with variable durations. (4) Conclusion: These results demonstrate the importance of experimental design on the assessment of the performance of automatic ARS classification algorithms. Furthermore, they also indicate, unlike what is stated in the literature, that the automatic classification of ARS is not a solved problem, as the algorithms’ performance decreases substantially under complex evaluation scenarios.

Analysis of Tracheal and Pulmonary Continuous Adventitious Respiratory Sounds in Asthma.

Continuous adventitious sounds (CAS) are commonly observed in obstructive pulmonary diseases and are of great clinical interest. However, their evaluation is generally subjective. We have previously developed an automatic CAS segmentation and classification algorithm for CAS recorded on the chest surface. The aim of this study is to establish whether these pulmonary CAS can be identified in a similar way using a tracheal microphone. Respiratory sounds were originally recorded from 25 participants using five contact microphones, four on the chest and one on the trachea, during three progressive respiratory maneuvers. In this work CAS component detection was performed on the tracheal channel using our automatic algorithm based on the Hilbert spectrum. The tracheal CAS detected were then compared to the previously analyzed pulmonary CAS. The sensitivity of CAS identification was lower at the tracheal microphone, with CAS that appeared simultaneously in all four pulmonary recordings more likely to be identified in the tracheal recordings. These observations could be due to the CAS being obscured by the lower SNR present in the tracheal recordings or not being transmitted through the airways to the trachea. Further work to optimize the algorithm for the tracheal recordings will be conducted in the future.

An open access database for the evaluation of respiratory sound classification algorithms

Objective: Over the last few decades, there has been significant interest in the automatic analysis of respiratory sounds. However, currently there are no publicly available large databases with which new algorithms can be evaluated and compared. Further developments in the field are dependent on the creation of such databases. Approach: This paper describes a public respiratory sound database, which was compiled for an international competition, the first scientific challenge of the IFMBE’s International Conference on Biomedical and Health Informatics. The database includes 920 recordings acquired from 126 participants and two sets of annotations. One set contains 6898 annotated respiratory cycles, some including crackles, wheezes, or a combination of both, and some with no adventitious respiratory sounds. In the other set, precise locations of 10 775 events of crackles and wheezes were annotated. Main results: The best system that participated in the challenge achieved an average score of 52.5% with the respiratory cycle annotations and an average score of 91.2% with the event annotations. Significance: The creation and public release of this database will be useful to the research community and could bring attention to the respiratory sound classification problem.

A Smartphone-Based System for Automated Bedside Detection of Crackle Sounds in Diffuse Interstitial Pneumonia Patients.

In this work, we present a mobile health system for the automated detection of crackle sounds comprised by an acoustical sensor, a smartphone device, and a mobile application (app) implemented in Android. Although pulmonary auscultation with traditional stethoscopes had been used for decades, it has limitations for detecting discontinuous adventitious respiratory sounds (crackles) that commonly occur in respiratory diseases. The proposed app allows the physician to record, store, reproduce, and analyze respiratory sounds directly on the smartphone. Furthermore, the algorithm for crackle detection was based on a time-varying autoregressive modeling. The performance of the automated detector was analyzed using: (1) synthetic fine and coarse crackle sounds randomly inserted to the basal respiratory sounds acquired from healthy subjects with different signal to noise ratios, and (2) real bedside acquired respiratory sounds from patients with interstitial diffuse pneumonia. In simulated scenarios, for fine crackles, an accuracy ranging from 84.86% to 89.16%, a sensitivity ranging from 93.45% to 97.65%, and a specificity ranging from 99.82% to 99.84% were found. The detection of coarse crackles was found to be a more challenging task in the simulated scenarios. In the case of real data, the results show the feasibility of using the developed mobile health system in clinical no controlled environment to help the expert in evaluating the pulmonary state of a subject.

Detection of Respiratory Crackle Sounds via an Android Smartphone-based System.

Pulmonary auscultation with traditional stethoscope, although useful, has limitations for detecting discontinuous adventitious respiratory sounds (crackles) that commonly occur in respiratory diseases. In this work, we present the development of a mobile health system for the automated detection of crackle sounds, comprised by an acoustical sensor, a smart phone device, and a mobile application (app) implemented in Android. The app allows the physician to record, store, reproduce, and analyze respiratory sounds directly on the smart phone. The algorithm for crackle detection was based on a time-varying autoregressive modeling. Performance of the automated detector was analyzed using synthetic fine and coarse crackle sounds randomly added to the basal respiratory sounds acquired from healthy subjects with different signal to noise ratios. Accuracy and sensitivity were found to range from 90.7% to 94.0% and from 91.2% to 94.2%, respectively. Application of the proposed mobile system to real acquired data from a patient with pulmonary fibrosis is also exemplified.

Enhancing our understanding of computerised adventitious respiratory sounds in different COPD phases and healthy people.

Timely diagnosis of acute exacerbations of COPD (AECOPD) is challenging as it depends on patients' reports. AECOPD are characterised by increased airway obstruction, mucus and air trapping, which results in changes in lung acoustics. Thus, adventitious respiratory sounds (ARS) may be useful to detect/monitor AECOPD. To evaluate computerised ARS changes during AECOPD. 25 non-hospitalised patients with AECOPD (16♂, 70 [62.5-77.0]yrs, FEV1 59 [31.5-73.0]%predicted) and 34 healthy volunteers (17♂, 63.5 [57.7-72.3]yrs, FEV1 103.0 [88.8-125.3]%predicted) were enrolled. ARS at anterior and posterior right and left chest were recorded at hospital presentation (T1), 15 days (T2) and 45 days (T3) after hospital presentation from patients with AECOPD and only once from healthy participants. A subsample of 9 patients (7♂; 66 [60.0-76.0]yrs; FEV1 62 [26.5-74.0]%predicted) was also included to study ARS pre-AECOPD (T0). Number of crackles and wheeze occupation rate (%Wh) were processed using validated algorithms. During AECOPD, patients presented more inspiratory crackles at T1 than T3 (p = 0.013) and more inspiratory %Wh at T1 than T2 (p = 0.006), at posterior chest. Patients with stable COPD presented more inspiratory crackles (p = 0.012), at posterior chest, and more expiratory %Wh, both at anterior (p < 0.001) and posterior (p = 0.001) chest, than healthy participants. No differences were observed for the remaining ARS parameters or subsamples (p > 0.05). Inspiratory crackles seem to persist until 15 days post exacerbation whilst inspiratory %Wh decreased after this period. ARS seem to be sensitive to monitor AECOPD. This information may allow advances in monitoring the recovery time of patients with AECOPD across all clinical and non-clinical settings.

Localization of adventitious respiratory sounds.

In a recent publication by Henry and Royston [J. Acoust. Soc. Am. 142, 1774-1783 (2017)], an algorithm was introduced to calculate the acoustic response to externally introduced and endogenous respiratory sounds within a realistic, patient-specific subglottal airway tree. This work is extended using an efficient numerical boundary element (BE) approach to calculate the resulting radiated sound field from the airway tree into the lung parenchyma taking into account the surrounding chest wall. Within the BE model of the left lung parenchyma, comprised of more than 6000 triangular surface elements, more than 30 000 monopoles are used to approximate complex airway-originated acoustic sources. The chest wall is modeled as a boundary condition on the parenchymal surface. Several cases were simulated, including a bronchoconstricted lung that had an internal acoustic source introduced in a bronchiole, approximating a wheeze. An acoustic source localization algorithm coupled to the BE model estimated the wheeze source location to within a few millimeters based solely on the acoustic field at the surface. Improved noninvasive means of locating adventitious respiratory sounds may enhance an understanding of acoustic changes correlated to pathology, and potentially provide improved noninvasive tools for the diagnosis of pulmonary diseases that uniquely alter acoustics.

Localization of adventitious respiratory sounds

Localization of adventitious respiratory sounds

Automatic adventitious respiratory sound analysis: A systematic review.

BackgroundAutomatic 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.ObjectiveTo 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 sourcesA 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 selectionOnly 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 extractionInvestigators 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 synthesisA 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.LimitationsDirect 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.ConclusionA 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.

Computerized respiratory sound analysis in people with dementia: a first-step towards diagnosis and monitoring of respiratory conditions

Computerized respiratory sound analysis has been shown to be an objective and reliable way to assess respiratory diseases. However, its application in non-collaborative populations, such as people with dementia, is still unknown. Therefore this study aimed to characterize normal and adventitious respiratory sounds (NRS; ARS) in older people with and without dementia.A cross-sectional study including two groups of 30 subjects with dementia and 30 subjects without dementia was performed. Digital auscultation was used to record NRS and ARS per breathing-phase (inspiration/expiration) at trachea and thorax. Frequency at percentiles 25, 50 and 75, frequency at maximum-intensity, maximum-intensity (Imax) and mean-intensity (Imean) characterized NRS. Crackle number, frequency, initial-deflection-width, 2cycle-duration, and largest-deflection-width and wheeze number, frequency and occupation-rate characterized ARS.Groups were similar in socio-demographics, except for anthropometrics. No significant differences were found between groups in NRS frequency or ARS at trachea or thorax. Significant lower Imax (inspiration: 36.88(29.42;39.92) versus 39.84(36.50;44.17) p = 0.007; expiration: 34.51(32.06;38.87) versus 42.33(36.92;44.98) p < 0.001) and Imean (inspiration: 15.23(12.08;18.60) versus 18.93(15.64;21.82) p = 0.003 and expiration: 14.57(12.08;18.30) versus 18.87(15.64;21.44) p = 0.001) at trachea and higher Imean (inspiration: 17.29(16.04;19.31) versus 16.45(15.05; 18.79) p = 0.005 and expiration: 16.71(15.31;18.56) versus 16.38(14.40;17.85) p = 0.011) at thorax were found in subjects with dementia when compared with subjects without dementia.To conclude, people with and without dementia had similar NRS and ARS characteristics, except for NRS intensity. Computerized respiratory sound analysis was feasible in a non-collaborative population. Further research is needed to enhance the use of respiratory acoustics in non-collaborative populations, with strong potential to be applied in different settings for diagnosis and monitoring purposes.

Integrated Approach for Automatic Crackle Detection Based on Fractal Dimension and Box Filtering

Crackles are adventitious respiratory sounds (RS) that provide valuable information on different respiratory conditions. Nevertheless, crackles automatic detection in RS is challenging, mainly when collected in clinical settings. This study aimed to develop an algorithm for automatic crackle detection/characterisation and to evaluate its performance and accuracy against a multi-annotator gold standard. The algorithm is based on 4 main procedures: i) recognition of a potential crackle; ii) verification of its validity; iii) characterisation of crackles parameters; and iv) optimisation of the algorithm parameters. Twenty-four RS files acquired in clinical settings were selected from 10 patients with pneumonia and cystic fibrosis. The algorithm performance was assessed by comparing its results with a multi-annotator gold standard agreement. High level of overall performance (F-score=92%) was achieved. The results highlight the potential of the algorithm for automatic crackle detection and characterisation of RS acquired in clinical settings.

Automatic Differentiation of Normal and Continuous Adventitious Respiratory Sounds Using Ensemble Empirical Mode Decomposition and Instantaneous Frequency

Differentiating normal from adventitious respiratory sounds (RS) is a major challenge in the diagnosis of pulmonary diseases. Particularly, continuous adventitious sounds (CAS) are of clinical interest because they reflect the severity of certain diseases. This study presents a new classifier that automatically distinguishes normal sounds from CAS. It is based on the multiscale analysis of instantaneous frequency (IF) and envelope (IE) calculated after ensemble empirical mode decomposition (EEMD). These techniques have two major advantages over previous techniques: high temporal resolution is achieved by calculating IF-IE and a priori knowledge of signal characteristics is not required for EEMD. The classifier is based on the fact that the IF dispersion of RS signals markedly decreases when CAS appear in respiratory cycles. Therefore, CAS were detected by using a moving window to calculate the dispersion of IF sequences. The study dataset contained 1494 RS segments extracted from 870 inspiratory cycles recorded from 30 patients with asthma. All cycles and their RS segments were previously classified as containing normal sounds or CAS by a highly experienced physician to obtain a gold standard classification. A support vector machine classifier was trained and tested using an iterative procedure in which the dataset was randomly divided into training (65%) and testing (35%) sets inside a loop. The SVM classifier was also tested on 4592 simulated CAS cycles. High total accuracy was obtained with both recorded (94.6% ± 0.3%) and simulated (92.8% ± 3.6%) signals. We conclude that the proposed method is promising for RS analysis and classification.

Automatic Crackle Detection Algorithm Based on Fractal Dimension and Box Filtering

Abstract Crackles are adventitious respiratory sounds that provide valuable information on different respiratory conditions. Crackles automatic detection in a respiratory sound file is challenging, and thus different signal processing methodologies have been proposed. However, limited testing of such methodologies, namely in respiratory sound files collected in clinical settings, has been conducted. This study aimed to develop an algorithm for automatic crackle detection and characterisation and to evaluate its performance and accuracy against a multi-annotator gold standard. The algorithm is based on three main procedures: i) extraction of a window of interest of a potential crackle (based on fractal dimension and box filtering techniques); ii) verification of the validity of the potential crackle considering computerised respiratory sound analysis established criteria; and iii) characterisation and extraction of crackle parameters. Twenty four 10-second files, acquired in clinical settings, were selected from 10 patients with pneumonia and cystic fibrosis. The algorithm performance was assessed by comparing its results with gold standard annotations (obtained by the agreement among three experts). A set of 7 parameters was optimised. High levels of sensitivity (SE=89%), positive predictive value (PPV=95%) and overall performance (F index=92%) were achieved. This promising result highlights the potential of the algorithm for automatic crackle's detection/characterisation in respiratory sounds acquired in clinical settings.

Computerized Adventitious Respiratory Sounds as Outcome Measures for Respiratory Therapy: A Systematic Review

There is a need to develop simple, noninvasive, and sensitive outcome measures for respiratory therapy. Adventitious respiratory sounds (ie, crackles and wheezes) can be objectively characterized with computerized respiratory sound analysis (CORSA) and have been shown to contribute for diagnosis purposes; however, their potential for use as outcome measures is unknown. Thus, this systematic review synthesizes the evidence on the use of computerized adventitious respiratory sounds as outcome measures. The Web of Knowledge, MEDLINE, EMBASE, and SCOPUS databases were searched. Reviewers independently selected studies according to the eligibility criteria. Effect sizes and 95% CIs were computed. Twelve studies with different designs (observational, n = 3; quasi-experimental n = 7; and randomized controlled trial, n = 2) were included. Eight studies were conducted with adults, and 4 studies with children. Most studies explored only one type of adventitious respiratory sound. For wheezes, the occupation rate seemed to be the most promising parameter to be used as an outcome measure, with high/medium effect sizes (0.62-1.82). For crackles, the largest deflection width showed high effect sizes (1.31 and 1.04); however, this was explored in only one study. Crackle number and 2-cycle duration presented conflicting information, with high/poor effect sizes depending on the study. Specific variables of each adventitious respiratory sound detected and characterized by CORSA showed high effect sizes and, thus, the potential to be used as outcome measures. Further research with robust study designs and larger samples (both of children and adult populations), and following CORSA guidelines is needed to build evidence-based knowledge on this topic.