Respiratory sounds. Advances beyond the stethoscope

Acoustical flow estimation in patients with obstructive sleep apnea during sleep

Tracheal respiratory sound analysis is a simple and non-invasive way to study the pathophysiology of the upper airways; it has recently been used for acoustical flow estimation and sleep apnea diagnosis. However in none of the previous studies, the accuracy of acoustical flow estimation was investigated neither during sleep nor in people with obstructive sleep apnea (OSA). In this study, we recorded tracheal sound, flow rate and head position from 11 individuals with OSA during sleep and wakefulness. We investigated two approaches for calibrating the parameters of acoustical flow estimation model based on the known data recorded during wakefulness and sleep. The results show that the acoustical flow estimation parameters change from wakefulness to sleep. Therefore, if the model is calibrated based on the data recorded during wakefulness, although the estimated flow follows the relative variations of the recorded flow, the quantitative flow estimation error would be high during sleep. On the other hand, when the calibration parameters are extracted from tracheal sound and flow recordings during sleep, the flow estimation error is less than 5%. These results confirm the reliability of acoustical methods for estimating breathing flow during sleep and detecting the partial or complete obstructions of the upper airways during sleep.

Respiratory flow-sound relationship during both wakefulness and sleep and its variation in relation to sleep apnea

Tracheal respiratory sound analysis is a simple and non-invasive way to study the pathophysiology of the upper airway and has recently been used for acoustic estimation of respiratory flow and sleep apnea diagnosis. However in none of the previous studies was the respiratory flow-sound relationship studied in people with obstructive sleep apnea (OSA), nor during sleep. In this study, we recorded tracheal sound, respiratory flow, and head position from eight non-OSA and 10 OSA individuals during sleep and wakefulness. We compared the flow-sound relationship and variations in model parameters from wakefulness to sleep within and between the two groups. The results show that during both wakefulness and sleep, flow-sound relationship follows a power law but with different parameters. Furthermore, the variations in model parameters may be representative of the OSA pathology. The other objective of this study was to examine the accuracy of respiratory flow estimation algorithms during sleep: we investigated two approaches for calibrating the model parameters using the known data recorded during either wakefulness or sleep. The results show that the acoustical respiratory flow estimation parameters change from wakefulness to sleep. Therefore, if the model is calibrated using wakefulness data, although the estimated respiratory flow follows the relative variations of the real flow, the quantitative flow estimation error would be high during sleep. On the other hand, when the calibration parameters are extracted from tracheal sound and respiratory flow recordings during sleep, the respiratory flow estimation error is less than 10%.

Respiratory wheeze detection system

Respiratory sound is associated with many lung diseases. By observing respiratory sound symptoms, we can know more about lung conditions. In this research, we construct an efficient lung sound recording system according to CORSA, and develop a spectrogram process flow technique to object wheeze. It is a low cost and efficient system. In clinic test, we also can precisely objective wheeze up to about 89%.

Vibration response imaging: a novel noninvasive tool for evaluating the initial therapeutic effect of noninvasive positive pressure ventilation in patients with acute exacerbation of chronic obstructive pulmonary disease

BACKGROUND

The popular methods for evaluating the initial therapeutic effect (ITE) of noninvasive positive pressure ventilation (NPPV) can only roughly reflect the therapeutic outcome of a patient's ventilation because they are subjective, invasive and time-delayed. In contrast, vibration response imaging (VRI) can monitor the function of a patient's ventilation over the NPPV therapy in a non-invasive manner. This study aimed to investigate the value of VRI in evaluating the ITE of NPPV for patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD).

METHODS

Thirty-six AECOPD patients received VRI at three time points: before NPPV treatment (T1), at 15 min of NPPV treatment (T2), and at 15 min after the end of NPPV treatment (T4). Blood gas analysis was also performed at T1 and at 2 hours of NPPV treatment (T3). Thirty-nine healthy volunteers also received VRI at T1 and T2. VRI examination at the time point T2 in either the patients or volunteers did not require any interruption of the on-going NPPV. The clinical indices at each time point were compared between the two groups. Moreover, correlations between the PaCO2 changes (T3 vs T1) and abnormal VRI scores (AVRIS) changes (T2 vs T1) were analyzed.

RESULTS

No significant AVRIS differences were found between T1 and T2 in the healthy controls (8.51 ± 3.36 vs. 8.53 ± 3.57, P > 0.05). The AVRIS, dynamic score, MEF score and EVP score showed a significant decrease in AECOPD patients at T2 compared with T1 (P < 0.05), but a significant increase at T4 compared with T2 (P < 0.05). We also found a positive correlation (R2 = 0.6399) between the PaCO2 changes (T3 vs T1) and AVRIS changes (T2 vs T1).

CONCLUSIONS

VRI is a promising noninvasive tool for evaluating the initial therapeutic effects of NPPV in AECOPD patients and predicting the success of NPPV in the early stage.

Thoracic auscultation in captive bottlenose dolphins (Tursiops truncatus), California sea lions (Zalophus californianus), and South African fur seals (Arctocephalus pusillus) with an electronic stethoscope

Thoracic auscultation is an important diagnostic method used in cases of suspected pulmonary disease in many species, as respiratory sounds contain significant information on the physiology and pathology of the lungs and upper airways. Respiratory diseases are frequent in marine mammals and are often listed as one of their main causes of death. The aim of this study was to investigate and report baseline parameters for the electronic-mediated thoracic auscultation of one cetacean species and two pinniped species in captivity. Respiratory sounds from 20 captive bottlenose dolphins (Tursiops truncatus), 6 California sea lions (Zalophus californianus), and 5 South African fur seals (Arctocephalus pusillus) were recorded with an electronic stethoscope. The sounds were analyzed for duration of the respiratory cycle, adventitious sounds, and peak frequencies of recorded sounds during expiration and inspiration as well as for sound intensity as reflected by waveform amplitude during the respiratory cycle. In respiratory cycles of the bottlenose dolphins' expiring "on command," the duration of the expiration was significantly shorter than the duration of the inspiration. In the examined pinnipeds of this study, there was no clear pattern concerning the duration of one breathing phase: Adventitious sounds were detected most often in bottlenose dolphins that were expiring on command and could be compared with "forced expiratory wheezes" in humans. This is the first report of forced expiratory wheezes in bottlenose dolphins; they can easily be misinterpreted as pathologic respiratory sounds. The peak frequencies of the respiratory sounds reached over 2,000 Hz in bottlenose dolphins and over 1,000 Hz in California sea lions and South African fur seals, but the variation of the frequency spectra was very high in all animals. To the authors' knowledge, this is the first systematic analysis of respiratory sounds of bottlenose dolphins and two species of pinnipeds.

Pendelluft in chronic obstructive lung disease measured with lung sounds

Objective. The phenomenon of pendelluft was described over five decades ago. In patients with regional variations in resistance and elastance, gas moves at the beginning of inspiration out of some alveoli into others. Gas moves in the opposite direction at the end of inspiration. The objective of this study was to apply the method of lung sounds mapping, which is known to provide regional information about gas flow, to study pendelluft in COPD patients. Methods. A 16-channel lung sound analyzer was used to collect sounds from patients with COPD (n = 90) and age-matched normals (n = 90). Pendelluft at the beginning of inspiration is expected to result in vesicular sounds leading the tracheal sound by a few milliseconds. Pendelluft at the end of inspiration is expected to result in vesicular sounds lagging the tracheal sound. These lead and lag times were calculated for the 14 chest wall sites. Results. The lead time was significantly longer in COPD patients: 123 ± 107 ms versus 48 ± 59 ms in controls (P < 0.0001). The lag time was also significantly longer in COPD patients: 269 ± 249 ms in COPD patients versus 147 ± 124 ms in controls (P < 0.0001). When normalized by the duration of the inspiration at the trachea, the lead was 14 ± 13% for COPD versus 4 ± 5% for controls (P < 0.0001). The lag was 28 ± 25% for COPD versus 13 ± 12% for controls (P < 0.0001). Both lead and lag correlated moderately with the GOLD stage (correlation coefficient 0.43). Conclusion. Increased lead and lag times in COPD patients are consistent with the phenomenon of pendelluft as has been observed by other methods.

Lung sound patterns help to distinguish congestive heart failure, chronic obstructive pulmonary disease, and asthma exacerbations

OBJECTIVES

Although congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), and asthma patients typically present with abnormal auscultatory findings on lung examination, respiratory sounds are not normally subjected to rigorous analysis. The aim of this study was to evaluate in detail the distribution of respiratory sound intensity in CHF, COPD, and asthma patients during acute exacerbation.

METHODS

Respiratory sounds throughout the respiratory cycle were captured and displayed using an acoustic-based imaging technique. Breath sound distribution was mapped to create a gray-scale sequence of two-dimensional images based on intensity of sound (vibration). Consecutive CHF (n = 22), COPD (n = 19), and asthma (n = 18) patients were imaged at the time of presentation to the emergency department (ED). Twenty healthy subjects were also enrolled as a comparison group. Geographical area of the images and respiratory sound patterns were quantitatively analyzed.

RESULTS

In healthy volunteers and COPD patients, the median (interquartile range [IQR]) geographical areas of the vibration energy images were similar, at 75.6 (IQR = 6.0) and 75.8 (IQR = 10.8) kilopixels, respectively (p > 0.05). Compared to healthy volunteers and COPD patients, areas for CHF and asthma patients were smaller, at 66.9 (IQR = 9.9) and 53.9 (IQR = 15.6) kilopixels, respectively (p < 0.05). The geographic area ratios between the left and right lungs for healthy volunteers and CHF and COPD patients were 1.0 (IQR = 0.2), 1.0 (IQR = 0.2), and 1.0 (IQR = 0.1), respectively. Compared to healthy volunteers, the geographic area ratio between the left and right lungs for asthma patients was 0.5 (IQR = 0.4; p < 0.05). In healthy volunteers and CHF patients, the ratios of vibration energy values at peak inspiration and expiration (peak I/E ratio) were 4.6 (IQR = 4.4) and 4.7 (IQR = 3.5). In marked contrast, the peak I/E ratios of COPD and asthma patients were 3.4 (= 2.1) and 0.1 (IQR = 0.3; p < 0.05), respectively.

CONCLUSIONS

The pilot data generated in this study support the concept that relative differences in respiratory sound intensity may be useful in distinguishing acute dyspnea caused by CHF, COPD, or asthma.

Investigating a compact phantom and setup for testing body sound transducers

Contact transducers are a key element in experiments involving body sounds. The characteristics of these devices are often not known with accuracy. There are no standardized calibration setups or procedures for testing these sensors. This study investigated the characteristics of a new computer-controlled sound source phantom for testing sensors. Results suggested that sensors with different sizes require special phantom requirements. The effectiveness of certain approaches on increasing the spatial and spectral uniformity of the phantom surface signal was studied. Non-uniformities > 20 dB were removable, which can be particularly helpful in comparing the characteristics of different size sensors more accurately.

Acoustics based assessment of respiratory diseases using GMM classification

The focus of this paper is to present a method utilizing lung sounds for a quantitative assessment of patient health as it relates to respiratory disorders. In order to accomplish this, applicable traditional techniques within the speech processing domain were utilized to evaluate lung sounds obtained with a digital stethoscope. Traditional methods utilized in the evaluation of asthma involve auscultation and spirometry, but utilization of more sensitive electronic stethoscopes, which are currently available, and application of quantitative signal analysis methods offer opportunities of improved diagnosis. In particular we propose an acoustic evaluation methodology based on the Gaussian Mixed Models (GMM) which should assist in broader analysis, identification, and diagnosis of asthma based on the frequency domain analysis of wheezing and crackles.

The effect of anthropometric variations on acoustical flow estimation: proposing a novel approach for flow estimation without the need for individual calibration

Tracheal sound average power is directly related to the breathing flow rate and recently it has attracted considerable attention for acoustical flow estimation. However, the flow-sound relationship is highly variable among people and it also changes for the same person at different flow rates. Hence, a robust model capable of estimating flow from tracheal sounds at different flow rates in a large group of individuals does not exist. In this paper, a model is proposed to estimate respiratory flow from tracheal sounds. The proposed model eliminates the dependence of the previous methods on calibrating the model for every individual and at different flow rates. To validate the model, it was applied to the respiratory sound and flow data of 93 healthy individuals. We investigated the statistical correlation between the model parameters and anthropometric features of the subjects. The results have shown that gender, height, and smoking are the most significant factors that affect the model parameters. Hence, we grouped nonsmoker subjects into four groups based on their gender and height. The average of model parameters in each group was defined as the group-calibrated model parameters. These models were applied to estimate flow from data of subjects within the same group and in the other groups. The results show that flow estimation error based on the group-calibrated model is less than 10%. The low estimation errors confirm the possibility of defining a general flow estimation model for subjects with similar anthropometric features with no need for calibrating the model parameters for every individual. This technique simplifies the acoustical flow estimation in general applications including sleep studies and patients' screening in health care facilities.

Automated analysis of crackles in patients with interstitial pulmonary fibrosis

Background. The crackles in patients with interstitial pulmonary fibrosis (IPF) can be difficult to distinguish from those heard in patients with congestive heart failure (CHF) and pneumonia (PN). Misinterpretation of these crackles can lead to inappropriate therapy. The purpose of this study was to determine whether the crackles in patients with IPF differ from those in patients with CHF and PN. Methods. We studied 39 patients with IPF, 95 with CHF and 123 with PN using a 16-channel lung sound analyzer. Crackle features were analyzed using machine learning methods including neural networks and support vector machines. Results. The IPF crackles had distinctive features that allowed them to be separated from those in patients with PN with a sensitivity of 0.82, a specificity of 0.88 and an accuracy of 0.86. They were separated from those of CHF patients with a sensitivity of 0.77, a specificity of 0.85 and an accuracy of 0.82. Conclusion. Distinctive features are present in the crackles of IPF that help separate them from the crackles of CHF and PN. Computer analysis of crackles at the bedside has the potential of aiding clinicians in diagnosing IPF more easily and thus helping to avoid medication errors.

Using quantitative breath sound measurements to predict lung function following resection

BACKGROUND

Predicting postoperative lung function is important for estimating the risk of complications and long-term disability after pulmonary resection. We investigated the capability of vibration response imaging (VRI) as an alternative to lung scintigraphy for prediction of postoperative lung function in patients with intrathoracic malignancies.

METHODS

Eighty-five patients with intrathoracic malignancies, considered candidates for lung resection, were prospectively studied. The projected postoperative (ppo) lung function was calculated using: perfusion scintigraphy, ventilation scintigraphy, and VRI. Two sets of assessments made: one for lobectomy and one for pneumonectomy. Clinical concordance was defined as both methods agreeing that either a patient was or was not a surgical candidate based on a ppoFEV1% and ppoDLCO% > 40%.

RESULTS

Limits of agreement between scintigraphy and VRI for ppo following lobectomy were -16.47% to 15.08% (mean difference = -0.70%;95%CI = -2.51% to 1.12%) and for pneumonectomy were -23.79% to 19.04% (mean difference = -2.38%;95%CI = -4.69% to -0.07%). Clinical concordance between VRI and scintigraphy was 73% for pneumonectomy and 98% for lobectomy. For patients who had surgery and postoperative lung function testing (n = 31), ppoFEV1% using scintigraphic methods correlated with measured postoperative values better than projections using VRI, (adjusted R2 = 0.32 scintigraphy; 0.20 VRI), however the difference between methods failed to reach statistical significance. Limits of agreement between measured FEV1% postoperatively and ppoFEV1% based on perfusion scintigraphy were -16.86% to 23.73% (mean difference = 3.44%;95%CI = -0.29% to 7.16%); based on VRI were -19.56% to 28.99% (mean difference = 4.72%;95%CI = 0.27% to 9.17%).

CONCLUSIONS

Further investigation of VRI as an alternative to lung scintigraphy for prediction of postoperative lung function is warranted.

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.

Classification of respiratory signals by linear analysis

The aim of this study is the classification of wheeze and non-wheeze epochs within respiratory sound signals acquired from patients with asthma and COPD. Since a wheeze signal, having a sinusoidal waveform, has a different behavior in time and frequency domains from that of a non-wheeze signal, the features selected for classification are kurtosis, Renyi entropy, f(50)/ f(90) ratio and mean-crossing irregularity. Upon calculation of these features for each wheeze and non-wheeze portion, the whole data scattered as two classes in four dimensional feature space is projected using Fisher Discriminant Analysis (FDA) onto the single dimensional space that separates the two classes best. Observing that the two classes are visually well separated in this new space, Neyman-Pearson hypothesis testing is applied. Finally, the correct classification rate is %95.1 for the training set, and leave-one-out approach pursuing the above methodology yields a success rate of %93.5 for the test set.

Vibration response imaging: evaluation of rater agreement in healthy subjects and subjects with pneumonia

Background

We evaluated pulmonologists variability in the interpretation of Vibration response imaging (VRI) obtained from healthy subjects and patients hospitalized for community acquired pneumonia.

Methods

The present is a prospective study conducted in a tertiary university hospital. Twenty healthy subjects and twenty three pneumonia cases were included in this study. Six pulmonologists blindly analyzed images of normal subjects and pneumonia cases and evaluated different aspects of VRI images related to the quality of data aquisition, synchronization of the progression of breath sound distribution and agreement between the maximal energy frame (MEF) of VRI (which is the maximal geographical area of lung vibrations produced at maximal inspiration) and chest radiography. For qualitative assessment of VRI images, the raters' evaluations were analyzed by degree of consistency and agreement.

Results

The average value for overall identical evaluations of twelve features of the VRI image evaluation, ranged from 87% to 95% per rater (94% to 97% in control cases and from 79% to 93% per rater in pneumonia cases). Inter-rater median (IQR) agreement was 91% (82-96). The level of agreement according to VRI feature evaluated was in most cases over 80%; intra-class correlation (ICC) obtained by using a model of subject/rater for the averaged features was overall 0.86 (0.92 in normal and 0.73 in pneumonia cases).

Conclusions

Our findings suggest good agreement in the interpretation of VRI data between different raters. In this respect, VRI might be helpful as a radiation free diagnostic tool for the management of pneumonia.

Accuracy of gray-scale coding in lung sound mapping

Stethoscope evaluation of the lungs is widely accepted and practiced; however, there are some widely recognized, major limitations with its use. A safe device that helped solve these limitations by translating sound into objective, quantifiable images would have clinical utility. Translating lung sounds into quantifiable images in which regional differences or asymmetry in intensities of breath sounds are presented as gradients in gray-scale is not a trivial process. Healthy lungs and lung pathology are characterized by different patterns of regional breath sound distribution and, therefore, the accuracy of mapping gray-scale images must be ensured in a controlled systematic fashion prior to clinical use of such a technique. Vibration response imaging (VRI) maps lung sounds from 40 sensors to a two-dimensional gray-scale image. To assess mapping accuracy, a simulated lung sound map with uniform signals was compared to modified maps where sound signals were reduced (1-3db) at one sensor. Also, 8 readers evaluated the gray-scale images. The computer algorithm accurately displayed gray-scale coding changes in correct locations in 97% of images. There was 95+/-4% accuracy rate by readers to correctly identify gray-scale changes. In addition, quantitative data at different stages of signal processing were investigated in a LSM of a subject with asthma. Signal processing was 97% accurate overall in that the gray-scale values from which the image was derived corresponded with intensity values from recorded signals. These results suggest VRI accurately maps acoustic signals to a gray-scale image and that trained readers can detect small changes.

Analysis of wheezes using wavelet higher order spectral features

Wheezes are musical breath sounds, which usually imply an existing pulmonary obstruction, such as asthma and chronic obstructive pulmonary disease (COPD). Although many studies have addressed the problem of wheeze detection, a limited number of scientific works has focused in the analysis of wheeze characteristics, and in particular, their time-varying nonlinear characteristics. In this study, an effort is made to reveal and statistically analyze the nonlinear characteristics of wheezes and their evolution over time, as they are reflected in the quadratic phase coupling of their harmonics. To this end, the continuous wavelet transform (CWT) is used in combination with third-order spectra to define the analysis domain, where the nonlinear interactions of the harmonics of wheezes and their time variations are revealed by incorporating instantaneous wavelet bispectrum and bicoherence, which provide with the instantaneous biamplitude and biphase curves. Based on this nonlinear information pool, a set of 23 features is proposed for the nonlinear analysis of wheezes. Two complementary perspectives, i.e., general and detailed, related to average performance and to localities, respectively, were used in the construction of the feature set, in order to embed trends and local behaviors, respectively, seen in the nonlinear interaction of the harmonic elements of wheezes over time. The proposed feature set was evaluated on a dataset of wheezes, acquired from adult patients with diagnosed asthma and COPD from a lung sound database. The statistical evaluation of the feature set revealed discrimination ability between the two pathologies for all data subgroupings. In particular, when the total breathing cycle was examined, all 23 features, but one, showed statistically significant difference between the COPD and asthma pathologies, whereas for the subgroupings of inspiratory and expiratory phases, 18 out of 23 and 22 out of 23 features exhibited discrimination power, respectively. This paves the way for the use of the wavelet higher order spectral features as an input vector to an efficient classifier. Apparently, this would integrate the intrinsic characteristics of wheezes within computerized diagnostic tools toward their more efficient evaluation.

Respiratory sound energy and its distribution patterns following clinical improvement of congestive heart failure: a pilot study

BACKGROUND

Although congestive heart failure (CHF) patients typically present with abnormal auscultatory findings on lung examination, respiratory sounds are not normally subjected to additional analysis. The aim of this pilot study was to examine respiratory sound patterns of CHF patients using acoustic-based imaging technology. Lung vibration energy was examined during acute exacerbation and after clinical improvement.

METHODS

Respiratory sounds throughout the respiratory cycle were captured using an acoustic-based imaging technique. Twenty-three consecutive CHF patients were imaged at the time of presentation to the emergency department and after clinical improvement. Digital images were created (a larger image represents more homogeneously distributed vibration energy of respiratory sound). Geographical area of the images and respiratory sound patterns were quantitatively analyzed. Data from the CHF patients were also compared to healthy volunteers.

RESULTS

The median (interquartile range) geographical areas of the vibration energy image of acute CHF patients without and with radiographically evident pulmonary edema were 66.9 (9.0) and 64.1(9.0) kilo-pixels, respectively (p < 0.05). After clinical improvement, the geographical area of the vibration energy image of CHF patients without and with radiographically evident pulmonary edema were increased by 18 +/- 15% (p < 0.05) and 25 +/- 16% (p < 0.05), respectively.

CONCLUSIONS

With clinical improvement of acute CHF exacerbations, there was more homogenous distribution of lung vibration energy, as demonstrated by the increased geographical area of the vibration energy image.

On arithmetic misconceptions of spectral analysis of biological signals, in particular respiratory sounds

Spectral analysis is one of the most common methods in sound signal analysis for approximating sound power. However, since the sound power is usually presented in logarithmic scale, it is important to consider the non-linearity effects of logarithm function. In this study, the misconceptions and implementation issues regarding noise power reduction and average power calculation are described. Respiratory sound analysis is utilized as an example to show these issues in a practical application. The results indicate that most of the errors happen during noise power reduction; they can be either due to substituting noise reduction by sound detection concept or/and representing the noise power in the very low frequency components instead of the signal power. Also, if the average powers of the signals are calculated in the wrong scale, the results do not represent the acoustical characteristics of the sounds; this is shown by considering the flow-sound relationship at different flow rates.

Formant analysis of breath and snore sounds

Formant frequencies of snore and breath sounds represent resonance in the upper airways; hence, they change with respect to the upper airway anatomy. Therefore, formant frequencies and their variations can be examined to distinguish between snore and breath sounds. In this paper, formant frequencies of snore and breath sounds are investigated and automatically grouped into 7 clusters based on K-Means clustering. First, formants clusters of breath and snore sounds of all subjects were investigated together and their union were calculated as the most probable ranges of the formants. The ranges for the first four formants which span the main frequency components of breath and snore sounds were found to be [20-400]Hz, [270-840]Hz, [500-1380]Hz and [910-1920]Hz. These ranges were then used as priori information to recalculate the formants of snore and breath sounds separately. Statistical t-test showed the 1(st) and 3(rd) formants to be the most characteristic features in distinguishing the breath and snore sounds from each other.

Base lung sound in diffuse interstitial pneumonia analyzed by linear and nonlinear techniques

Abnormal lung sounds in diffuse interstitial pneumonia have been characterized by the presence of crackles. However, few attempts have tried to analyze the sound where crackles are immersed. In this work base lung sounds (BLS) were analyzed by linear and nonlinear techniques to find possible differences between normal subjects and patients with diffuse interstitial pneumonia. In both groups, segments of lung sounds (crackles absent) and segments of BLS (lung sound in between crackles) were analyzed from acquired lung sounds from four points at the posterior chest, two apical and two basal. In this study, 8 healthy subjects and 8 patients participated and BLS were analyzed by spectral percentiles and sample entropy. Although spectral percentiles and sample entropy (SampEn) tended to be lower in the group of patients, statistical differences (p0.05) between normal subjects and patients were found in BLS at the left hemithorax at basal and apical regions, while at the right hemithorax significant differences were found only at the basal region using the nonlinear technique. We conclude that in respect to normal subjects, BLS of patients with diffuse interstitial pneumonia present differences as assessed by SampEn, so that BLS by itself provides useful information. Moreover, it seems that nonlinear technique did a better discrimination of abnormal BLS than spectral percentile parameters.

Comparison of flow-sound relationship for different features of tracheal sound

In recent years, respiratory flow estimation using tracheal sounds has received considerable attention. In this paper, four different features of tracheal sound are investigated and their relationships with flow at different target flow rates are examined during inspiration and expiration phases. The features include average power (AvgPwr), logarithm of the variance (LogVar), logarithm of the range (LogRng) and logarithm of the envelop (LogEnv) of tracheal sound. For each feature a linear model is fitted to the flow and the feature. The results show that LogVar is the best feature to describe flow-sound relationship with a linear model, while the slope of the linear model using AvgPwr shows the largest deviation from a line with changes in target flow rates. Also, the distance from origin of the linear model using any feature changes linearly with variations of target flow.

Respiratory acoustic thoracic imaging (RATHI): assessing intrasubject variability

Respiratory acoustic thoracic imaging (RATHI) permits analysing lung sounds (LS) temporal and spatial distribution, however, a deep understanding of RATHI repeatability associated with the pulmonary function is necessary. As a consequence, in the current work intrasubject variability of RATHI is evaluated at different airflows. For generating RATHIs, LS were acquired at the posterior thoracic surface. The associated image was computed at the inspiratory phases by interpolation through a Hermite function. The acoustic information of eleven subjects was considered at airflows of 1.0, 1.5 and 2.0 L/s. Several RATHIs were generated for each subject according to the number of acquired inspiratory phases. Quadratic mutual information based on Cauchy-Schwartz inequality (I(CS)) was used to evaluate the degree of similitude between intrasubject RATHIs. The results indicated that, for the same subject, I(CS) averaged 0.893, 0.897, and 0.902, for airflows of 1.0, 1.5, and 2 L/s, respectively. In addition, when the airflow was increased, increments in intensity values and in the dispersion of the spatial distribution reflected in RATHI were observed. In conclusion, since the intrasubject variability of RATHI was low for airflows between 1.0 and 2.0 L/s, the pattern of sound distribution during airflow variations is repeatable but differences in sound intensity should be considered.

Evaluation of obstructive lung disease with vibration response imaging

As optimal treatment and prognosis differ between asthma and COPD, a new diagnostic approach to differentiating between the two disorders would be clinically desirable. We evaluated the utility of vibration response imaging in differentiating between asthma and COPD. Sixty-six subjects with asthma or COPD were recorded, before and after the administration of a short-acting bronchodilator, using a computerized lung sound analysis device. Gray-scale images of breath sound distribution in the lungs, quantitative data in breath sound graphs (timing, amplitude) and automatic crackle and wheeze detection programs were used to differentiate between asthma and COPD subjects. Imaging data were compared with the clinical diagnosis, made by the standard methods (medical history, physical examination, and spirometric indices). Blinded evaluation of images demonstrated a significantly higher rate of improvement in image dynamics, shape and overall improvement following bronchodilator in subjects with asthma compared with those with COPD. Quantitative data showed distinct patterns in timing and amplitude for these two pathologies. Combined analyses based on qualitative image evaluation and quantitative data demonstrated an overall 85% accuracy (84% for asthma, 86% for COPD) in differentiating between asthma and COPD. Combined qualitative and quantitative evaluations of lung sounds are quite sensitive in distinguishing between lung sound recordings of COPD and asthma individuals. Lung sound features of synchronization in timing and intensity provide objective data that may further differentiate these two airway disorders.

Validation of automatic wheeze detection in patients with obstructed airways and in healthy subjects

Computerized lung-sound analysis is a sensitive and quantitative method to identify wheezing by its typical pattern on spectral analysis. We evaluated the accuracy of the VRI, a multi-sensor, computer-based device with an automated technique of wheeze detection. The method was validated in 100 sound files from seven subjects with asthma or chronic obstructive pulmonary disease and seven healthy subjects by comparison of auscultation findings, examination of audio files, and computer detection of wheezes. Three blinded physicians identified 40 sound files with wheezes and 60 sound files without wheezes. Sensitivity and specificity were 83% and 85%, respectively. Negative predictive value and positive predictive value were 89% and 79%, respectively. Overall inter-rater agreement was 84%. False positive cases were found to contain sounds that simulate wheezes, such as background noises with high frequencies or strong noises from the throat that could be heard and identified without a stethoscope. The present findings demonstrate that the wheeze detection algorithm has good accuracy, sensitivity, specificity, negative predictive value and positive predictive value for wheeze detection in regional analyses with a single sensor and multiple sensors. Results are similar to those reported in the literature. The device is user-friendly, requires minimal patient effort, and, distinct from other devices, it provides a dynamic image of breath sound distribution with wheeze detection output in less than 1 minute.

Pulmonary disease management system with distributed wearable sensors

A pulmonary disease management system with on-body and near-body sensors is introduced in this presentation. The system is wearable for continuous ambulatory monitoring. Distributed sensor data is transferred through a wireless body area network (BAN) to a central controller for real time analysis. Physiological and environmental parameters are monitored and analyzed using prevailing clinical guidelines for self-management of environmentally-linked pulmonary ailments. The system provides patients with reminders, warnings, and instructions to reduce emergency room and physician visits, and improve clinical outcomes.

Acoustic obstructive sleep apnea detection

Obstructive sleep apnea (OSA) is a common respiratory disorder during sleep, in which the airways are collapsed and impair the respiration. Apnea is s cessation of airflow to the lungs which lasts at least for 10s. The current gold standard method for OSA assessment is full night polysomnography (PSG); however, its high cost, inconvenience for patients and immobility have persuaded researchers to seek simple and portable devices to detect OSA. In this paper, we report on developing a new system for OSA detection and monitoring, which only requires two data channels: tracheal breathing sounds and the blood oxygen saturation level (S(a)O(2)). A fully automated method was developed that uses the energy of breathing sounds signals to segment the signals into sound and silent segments. Then, the sound segments are classified into breath, snore (if exists) and noise segments. The S(a)O(2) signal is analyzed to find the rises and drops in the S(a)O(2) signal. Finally, a fuzzy algorithm was developed to use this information and detect apnea and hypopnea events. The method was evaluated on the data of 40 patients simultaneously with full night PSG study, and the results were compared with those of the PSG. The results show high correlation (96%) between our system and PSG. Also, the method has been found to have sensitivity and specificity values of more than 90% in differentiating simple snorers from OSA patients.

Assessment of asymmetric lung disease in intensive care unit patients using vibration response imaging

BACKGROUND

Vibration response imaging (VRI) is a computer-based technology that creates a visual dynamic two-dimensional image of distribution of vibration within the lung during the respiratory process. The acoustic signals, recorded from 36 posteriorly positioned surface skin sensors, are transferred to a hardware board where several stages of filtering are applied to select a specific frequency band. The filtered output signal frequencies are presented as a gray-scale coded dynamic image, consisting of a series of 0.17 s frames, and as a table featuring the percentage contribution of each lung to the total vibration signal.

METHODS

We describe the VRI technology in detail and examine images obtained from consecutive intensive care unit (ICU) patients with one diseased lung on chest radiograph. ICU patients with normal chest radiographs are presented as controls. Analysis of the image was performed by comparing the weighted pixel count analysis from both lungs. In this method, the pixels in the image were assigned values based on their grayscale color with the darker pixels assigned higher values.

RESULTS

In patients with normal chest radiographs, the right and left lungs developed similarly in dynamic VRI images, and the percent lung vibrations from both sides were comparable (53%+/-12% and 47%+/-12%, respectively). In ICU patients with asymmetric lung disease, however, the percent lung vibrations from the diseased and nondiseased lungs were 27%+/-23% and 73%+/-23%, respectively (P<0.001). In patients with asymmetric lung disease (one lung has moderate to severe disease and the other appears normal or close to normal as per chest radiograph), the diseased lung usually appeared in VRI as irregular, smaller, and lighter in color (reduced vibration signal) when compared to the nonaffected lung. The weighted pixel count from diseased and nondiseased lungs were 33%+/-21% and 67%+/-21%, respectively (P < 0.003).

CONCLUSION

The VRI technology may provide a radiation-free method for identifying and tracking of asymmetric lung parenchymal processes.

Asynchrony between left and right lungs in acute asthma

BACKGROUND

Asthma is a disease of air flow obstruction. Transmitted sounds can be analyzed in detail and may shed light upon the physiology of asthma and how it changes over time. The goals of this study were to use a computerized analytic acoustic tool to evaluate respiratory sound patterns in asthmatic patients during acute attacks and after clinical improvement and to compare asthmatic profiles with those of normal individuals.

METHODS

Respiratory sound analysis throughout the respiratory cycle was performed on 22 symptomatic asthma patients at the time of presentation to the emergency department (ED) and after clinical improvement. Fifteen healthy volunteers were analyzed as a control group. Vibrations patterns were plotted. Right and left lungs were analyzed separately.

RESULTS

Asthmatic attacks were found to be correlated with asynchrony between lungs. In normal subjects, the inspiratory and expiratory vibration energy peaks (VEPs) occurred almost simultaneously in both lungs; the time interval between right and left expiratory VEPs was 0.006 +/- 0.012 seconds. In symptomatic asthmatic patients on admission, the time interval between right and left expiratory VEPs was 0.14 +/- 0.09 seconds and after clinical improvement the interval decreased to 0.04 +/- 0.04 seconds. Compared to healthy volunteers, asynchrony between two lungs was increased in asthmatics (p < 0.05). The asynchrony was significantly reduced after clinical improvement (p < 0.05).

CONCLUSIONS

Respiratory sound analysis demonstrated significant asynchrony between right and left lungs in asthma exacerbations, a finding which, to our knowledge, has never been reported to date. The asynchrony is significantly reduced with clinical improvement following treatment.

Monitoring of wheeze duration during spontaneous respiration in asthmatic patients

Respiratory sound analysis can offer important information related to pulmonary diseases. Wheezes have been reported as adventitious respiratory sounds in asthmatic or obstructive patients, during forced exhalation maneuvers. In this work, we propose a method for monitoring and analysis of respiratory sounds in frequency domain, during spontaneous ventilation. The database analyzed was acquired during spontaneous ventilation for 120 seconds (DBsv), of 26 asthmatics patients. Using an autoregressive model (AR, order 16), the Power Spectral Density (PSD) was calculated for every phase of expiration and inspiration and the maximum frequency (fp) was estimated. From this parameter we study the time duration of the wheezes. The effect of bronchodilator inhalation in asthmatic patients was studied analyzing the duration of the wheezes in the bandwidth 600-2000 Hz (HFband). The wheeze duration is evaluated as the number of consecutive segments, with fp is inside of HFband, (for 3 or more segments in a cycle). The difference of the wheeze duration inside the respiratory cycles (Dwheez), before and after bronchodilator inhalation (POST) was evaluated. It was found a good correlation between Dwheez and FEV 1% improvement (FEV 1%_imp), for FEV1%_imp greater than 8%, whereas values FEV1%_imp less than 8% did not show any change of Dwheez. This last result suggests no difference in the wheeze duration between the baseline and POST records. This method could predict the FEV1%_imp by means of estimation of Dwheez during spontaneous ventilation.

Computerized classification of normal and abnormal lung sounds by multivariate linear autoregressive model

This work proposes multichannel acquisition of lung sounds by a microphone array, feature extraction by a multivariate AR (MAR) model, dimensionality reduction of the feature vectors (FV) by SVD and PCA and, their classification by a supervised neural network. A microphone array of 25 sensors was attached on the thoracic surface of the subjects, who were breathing at 1.5 L/sec. The supervised neural network used the backpropagation learning algorithm based on the Levenberg-Marquardt rule. Figures of merit for the classification task by the neural net include the percentage of correct classification during training, testing and validation phases as well as sensitivity, specificity and performance. MAR in combination with PCA provided the best average percentage of correct classification with acoustic information not seen by the neural network during the training phase (87.68%). The results state the advantages of a microphone array for the classification of normal and abnormal acoustic pulmonary information in diffuse interstitial pneumonia and for this goal, the authors assume that not only the crackles and their number indicates the severity of the disease, but the basal respiratory signal could be also affected.

Analysis of respiratory sounds: state of the art

OBJECTIVE

This paper describes state of the art, scientific publications and ongoing research related to the methods of analysis of respiratory sounds.

METHODS AND MATERIAL

Review of the current medical and technological literature using Pubmed and personal experience.

RESULTS

The study includes a description of the various techniques that are being used to collect auscultation sounds, a physical description of known pathologic sounds for which automatic detection tools were developed. Modern tools are based on artificial intelligence and on technics such as artificial neural networks, fuzzy systems, and genetic algorithms…

CONCLUSION

The next step will consist in finding new markers so as to increase the efficiency of decision aid algorithms and tools.

Nonlinear characteristics of wheezes as seen in the wavelet higher-order spectra domain

The aim of this study was to capture and analyze the nonlinear characteristics of asthmatic wheezes, reflected in the quadrature phase coupling of their harmonics, as they evolve over time within the breathing cycle. To achieve this, the continuous wavelet transform was combined with third-order statistics/spectra. Wheezes from diagnosed asthmatic patients were drawn from a lung sound database and analyzed in the time-bi-frequency domain. The analysis results justified the efficient performance of this combinatory approach to reveal and quantify the evolution of wheeze nonlinearities with time.

Classifying respiratory sounds with different feature sets

In this study, different feature sets are used in conjunction with (k-nearest neighbors) k-NN and artificial neural network (ANN) classifiers to address the classification problem of respiratory sound signals. A comparison is made between the performances of k-NN and ANN classifiers with different feature sets derived from respiratory sound data acquired from one microphone placed on the posterior chest area. Each subject is represented by a single respiration cycle divided into sixty segments from which three different feature sets consisting of 6th order AR model coefficients, wavelet coefficients and crackle parameters in addition to AR model coefficients are extracted. Classification experiments are carried out on inspiration and expiration phases separately. The two class recognition problem between healthy and pathological subjects is addressed.