Time-Frequency Detection and Analysis of Wheezes During Forced Exhalation

Abstract

The objective of the present work was to detect and analyze wheezes by means of a highly sensitive time-frequency algorithm. Automatic measurements were compared with clinical auscultation for forced exhalation segments from 1.2 to 0 liters/second (l/s). Sensitivities between 100% and 71%, as a function of flow level related to wheezing segments detection, were achieved. Time-frequency wheeze parameters were measured for the flow range from 1.2 to 0.2 l/s. Wheezes were detected in both analyzed groups; asthmatics (N = 16) and control subjects (N = 15). Significant differences between groups were found for the mean number of wheezes detected at basal condition (p = 0.0003). Frequency parameter differences were also significant (0.0112 < p < 0.0307). All these parameters were also studied after applying a bronchodilator drug (Terbutaline). Significant differences between patient groups were found when studying the changes in the number of wheezes for each patient (p = 0.0195). Finally, limited bandwidth parameters, which measure the bronchodilator response, were also studied.

Wheeze Recognition Based on 2d Bilateral Filtering of Spectrogram

“…The algorithms combined with classification models provide more precise but slower wheezing detection. Recent attempts for achieving higher sensitivity and efficient detection performance include a set of criteria in the timefrequency domain [15][16][17]. These criteria refer to time duration, pitch range and magnitude of wheezes in their time-frequency representation by means of spectrogram analysis.…”

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

“…5,28 Algorithms based on fast Fourier transformation were the most used to automatically detect ARSs. Two studies 28,29 used an algorithm based on short-time Fourier transformation, and one study 25 used a modification of the algorithm proposed by Shabtai-Musih et al 31 and HomsCorbera et al 32 A total of 9 studies analyzed WHs (3 studies were conducted in children), 2 studies analyzed CRs, 5,27 and one study analyzed both WHs and CRs in children. 30 Two studies detected breathing cycles automatically; one study 5 used an analogous method reported by Qiu et al, 33 and the other study 30 used the algorithm of Huq and Moussavi.…”

Performance evaluation of the Hilbert–Huang transform for respiratory sound analysis and its application to continuous adventitious sound characterization

“…This study implies a step forward in the analysis of CAS, as most previous approaches for CAS analysis [3][4][5][6][7]9] mainly focused on differentiating CAS from normal RS, but not on analyzing CAS features, such as duration and mean frequency, which are the most relevant clinical parameters. Moreover, some of those studies used spectrogram for CAS detection and extraction.…”

“…In general, RS signals are comprised of normal RS and may contain superimposed abnormal RS, such as continuous (CAS) and the most commonly used and straightforward techniques for RS characterization. In CAS analysis, spectrogram has been the most widely used TFD [3][4][5][6][7], despite its poor and window-dependent resolution. Nevertheless, more advanced TFDs have recently been proposed for CAS analysis, either through combining wavelet decomposition with third order spectra features [8], or by deriving a temporal-spectral dominance spectrogram from the shorttime Fourier transform [9].…”