Assessing the variability in respiratory acoustic thoracic imaging (RATHI).

Multichannel analysis of lung sounds (LSs) has enabled the generation of a functional image for the temporal and spatial study of LS intensities in healthy and diseased subjects; this method is known as respiratory acoustic thoracic imaging (RATHI). This acoustic imaging technique has been applied to diverse pulmonary conditions, but it is important to contribute to the understanding of RATHI characteristics, such as acoustic spatial distribution, dependence on airflow and variability. The purpose of the current study is to assess the intra-subject and inter-subject RATHI variabilities in a cohort of 12 healthy male subjects (24.3±1.5 years) using diverse quantitative indices. The indices were obtained directly from the acoustic image and did not require scores from human raters, which helps to prevent inter-observer variability. To generate the acoustic image, LSs were acquired at 25 positions on the posterior thoracic surface by means of airborne sound sensors with a wide frequency band from 75 up to 1000 Hz under controlled airflow conditions at 1.0, 1.5 and 2.0 L/s. To assess intra-subject variability, the degree of similitude between inspiratory acoustic images was evaluated through quadratic mutual information based on the Cauchy-Schwartz inequality (I(CS)). The inter-subject variability was assessed by an image registration procedure between RATHIs and X-ray images to allow the computation of average and variance acoustic image in the same coordinate space. The results indicated that intra-subject RATHI similitude, reflected by I(CS-global), averaged 0.960±0.008, 0.958±0.008 and 0.960±0.007 for airflows of 1.0, 1.5, and 2L/s, respectively. As for the inter-subject variability, the variance image values for three airflow conditions indicated low image variability as they ranged from 0.01 to 0.04. In conclusion, the assessment of intra-subject and inter-subject variability by similitude indices indicated that the acoustic image pattern is repeatable along different respiratory cycles and across different subjects. Therefore, RATHI could be used to explore different aspects of spatial distribution and its association with regional pulmonary ventilation.

Assessment of multichannel lung sounds parameterization for two-class classification in interstitial lung disease patients.

This work deals with the assessment of different parameterization techniques for lung sounds (LS) acquired on the whole posterior thoracic surface for normal versus abnormal LS classification. Besides the conventional technique of power spectral density (PSD), the eigenvalues of the covariance matrix and both the univariate autoregressive (UAR) and the multivariate autoregressive models (MAR) were applied for constructing feature vectors as input to a supervised neural network (SNN). The results showed the effectiveness of the UAR modeling for multichannel LS parameterization, using new data, with classification accuracy of 75% and 93% for healthy subjects and patients, respectively.

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.

Asymmetry in lung sound intensities detected by respiratory acoustic thoracic imaging (RATHI) and clinical pulmonary auscultation.

RATHI was introduced as an attempt to further improve the association between anatomical zones and specific breathing activity, both spatially and temporally. This work compares RATHI with clinical pulmonary auscultation (PA) to assess the concordance between both procedures to detect asymmetries in lung sound (LS) intensities. Twelve healthy young males participated in the study and were auscultated by two experts. RATHI consisted in the acquisition of acoustical signals with an array of 5x5 sensors, while experts auscultated and described the intensity of LS heard using the same stethoscope on each sensor's position within the array. Comparisons were established looking for intensity asymmetries between apical vs. basal pulmonary regions and right vs. left hemithorax. By RATHI, most of the subjects showed asymmetries between apical and basal regions higher than 20%, whereas between left and right hemithorax asymmetries higher than 20% occurred only half of the time. RATHI and PA agreed 83 to 100% when apical to base acoustical information was compared, but when left to right asymmetries were considered these figures were about 40 to 50%. We concluded that RATHI has advantages as it gave more detailed and measurable information on LS than clinicians, who could not detect intensity asymmetries mainly below 20%.

Imaging of simulated crackle sounds distribution on the chest.

Crackles sounds have been associated with several pulmonary pathologies and diverse algorithms have been proposed for extracting and counting them from the acquired lung sound. These tasks depend among other factors, of the relation between the magnitude of the crackle and the background lung sound. In this work, we explore multivariate signal processing to deal with the tasks and propose a new concept, the discontinuous adventitious sounds imaging. The image formation is founded on the results of two proposed methodologies that use an autoregressive (AR) model. In the first case, the AR coefficients feed an artificial neural network (ANN) to classify temporal acoustic information as healthy or sick and; in the second case, a time-variant AR (TVAR) model, obtained by the RLS algorithm, permits to detect changes in the TVAR coefficients to be associated with the number of crackles. For AR-ANN, the ratio of the temporal windows classified as sick to the classified as healthy is used as an index to form the adventitious image, while for TVAR-RLS, an estimation of the number of crackles is obtained to form the corresponding image. The results indicated that fine and coarse crackles could be detected and counted even with very low crackle magnitude so that the formation of a crackle distribution image was consistent.

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.

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.

Acoustic thoracic images for transmitted glottal sounds.

Sound transmission has been of interest for many years in an attempt to study the structure of the lung and different researches have shown that artificial sounds produce a lateralization of sound information at the thoracic surface. Most of these studies have use non-simultaneous recording and input sounds introduced at the mouth or other thoracic points. In this paper, we present acoustic thoracic images, for transmitted glottal sounds, formed by a multichannel system with an array of 5x5 microphones. The study was done using 4 healthy subjects and 4 subjects having diffuse interstitial pneumonia. In both groups of subjects, it was found that the thorax behaves as a lowpass filter depending on the physical properties of its components, and that the transmitted acoustic thoracic imaging (TATHI) could reflect such properties. In most of the healthy subjects right to left asymmetries and heterogeneous apical to basal distribution were found. In patients the lateral dominance was lost and an intensity increment in the frequency band of 300 to 600 Hz was revealed. We conclude that TATHI permits to observe easily the spatial extension of the disease through sound transmission.

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.

Respiratory acoustic thoracic imaging (RATHI): assessing deterministic interpolation techniques.

As respiratory sounds contain mechanical and clinical pulmonary information, technical efforts have been devoted during the past decades to analysing, processing and visualising them. The aim of this work was to evaluate deterministic interpolating functions to generate surface respiratory acoustic thoracic images (RATHIs), based on multiple acoustic sensors. Lung sounds were acquired from healthy subjects through a 5 x 5 microphone array on the anterior and posterior thoracic surfaces. The performance of five interpolating functions, including the linear, cubic spline, Hermite, Lagrange and nearest neighbour method, were evaluated to produce images of lung sound intensity during both breathing phases, at low (approximately 0.5ls(-1)) and high (approximately 1.0ls(-1)) airflows. Performance indexes included the normalised residual variance nrv (i.e. inaccuracy), the prediction covariance cv (i.e. precision), the residual covariance rcv (i.e. bias) and the maximum squared residual error semax (i.e. tolerance). Among the tested interpolating functions and in all experimental conditions, the Hermite function (nrv=0.146 +/- 0.059, cv= 0.925 +/- 0.030, rcv = -0.073 +/- 0.068, semax = 0.005 +/- 0.004) globally provided the indexes closest to the optimum, whereas the nearest neighbour (nrv=0.339 +/- 0.023, cv = 0.870 +/- 0.033, rcv= 0.298 +/- 0.032, semax = 0.007 +/- 0.005) and the Lagrange methods (nrv = 0.287 +/- 0.148, cv = 0.880 +/- 0.039, rcv = -0.524 +/- 0.135, semax = 0.007 +/- 0.0001) presented the poorest statistical measurements. It is concluded that, although deterministic interpolation functions indicate different performances among tested techniques, the Hermite interpolation function presents a more confident deterministic interpolation for depicting surface-type RATHI.

Gas exchange at rest during simulated altitude in patients with chronic lung disease.

BACKGROUND: To characterize the gasometric and oximetric response to simulated altitudes of 3,100 m and sea level of patients with Chronic Obstructive Pulmonary Disease (COPD) and Interstitial Lung Disease (ILD) studied at 2,240 m above sea level. METHODS: Consecutive stable patients with COPD and ILD were studied at the National Institute of Respiratory Diseases, a referral center for pulmonary diseases in Mexico City, and a healthy control group. The patients breathed room air (FIO2 = 0.21), for at least 15 min, then, a hypoxic mixture (FIO2 = 0.18, simulating 3,100 m), and finally, a hyperoxic mixture (FIO2 = 0.28, simulating sea level). Arterial blood gases and oxygen saturation were measured by a pulse oximeter at the end of each stage. RESULTS: Twelve patients with COPD, 13 patients with ILD and 11 healthy controls were studied. The PaCO2 and pH were constant in the three study stages in both groups of patients and controls. A slope of PaO2 vs. altitude of 9 Torr per Km was found for each of the study's patients, either by simple linear regression or multiple regression, which is identical to that previously obtained at sea level with COPD patients (Gong et al.). Oxygen desaturation per Km of altitude change was alinear, higher for the hypoxic than for the hyperoxic challenge and more severe for the most hypoxic patients. CONCLUSIONS: Exposure tests to simulated altitudes are safe, and orient the physician concerning the patient's condition at altitudes different from the place where the measurement is done. Alveolar ventilation remains constant despite hypoxia or hyperoxia during the challenges. A computer model of the lung reproduces many of the findings in the challenges of this study.