Date Published: March 12, 2019
Publisher: Public Library of Science
Author(s): Renard Xaviero Adhi Pramono, Syed Anas Imtiaz, Esther Rodriguez-Villegas, Saeid Ghavami.
Chronic Respiratory Diseases (CRDs), such as Asthma and Chronic Obstructive Pulmonary Disease (COPD), are leading causes of deaths worldwide. Although both Asthma and COPD are not curable, they can be managed by close monitoring of symptoms to prevent worsening of the condition. One key symptom that needs to be monitored is the occurrence of wheezing sounds during breathing since its early identification could prevent serious exacerbations. Since wheezing can happen randomly without warning, a long-term monitoring system with automatic wheeze detection could be extremely helpful to manage these respiratory diseases. This study evaluates the discriminatory ability of different types of feature used in previous related studies, with a total size of 105 individual features, for automatic identification of wheezing sound during breathing. A linear classifier is used to determine the best features for classification by evaluating several performance metrics, including ranksum statistical test, area under the sensitivity-–specificity curve (AUC), F1 score, Matthews Correlation Coefficient (MCC), and relative computation time. Tonality index attained the highest effect size, at 87.95%, and was found to be the feature with the lowest p-value when ranksum significance test was performed. Third MFCC coefficient achieved the highest AUC and average optimum F1 score at 0.8919 and 82.67% respectively, while the highest average optimum MCC was obtained by the first coefficient of a 6th order LPC. The best possible combination of two and three features for wheeze detection is also studied. The study concludes with an analysis of the different trade-offs between accuracy, reliability, and computation requirements of the different features since these will be highly useful for researchers when designing algorithms for automatic wheeze identification.
Chronic Respiratory Diseases (CRDs) affect over 15% of the world population. It is estimated that more than 235 million people suffer from asthma worldwide, with the disease causing in excess off 300,000 deaths per year . The prevalence of COPD is even higher with more than 250 million cases reported annually, further resulting in over 3 million deaths globally . COPD is also predicted to become the third leading cause of deaths worldwide by 2030, just behind ischaemic heart disease and cerebrovascular disease . Furthermore, the economic impact of these respiratory diseases is also very high. As an illustration, in the UK they cost the National Health Service approximately 3 and 2 million GBP, for asthma and COPD, respectively .
To assess and compare the performance of different features to distinguish between wheeze and normal respiratory events, a comprehensive review of the existing works in literature was performed. The review focused mainly on studies which carried out classification between wheeze and normal respiratory events. Based on this, a number of different features were selected, with the inclusion criteria being that enough details about a feature were available for its implementation to discriminate between wheeze and normal respiratory events using a simple linear threshold. The details about the selected features, preprocessing stages, and classification, are discussed in this section.
This paper evaluated the performance of different features for automated detection of wheezes from respiratory sounds. The top performing features were determined using different objective functions and performance metrics. The rational for doing this was that the best feature is usually different in various applications and is heavily dependent on a number of constraints resulting in a number of trade-offs. For example, features with better F1 score would be more useful when the trade-off between sensitivity and number of false positives is important. Similarly, a higher MCC value for a feature generally represents a more balanced overall performance.
This paper has presented a comprehensive evaluation of the discriminatory abilities of different types of time, spectral, wavelet, and cepstral features with a total size of 105 for automatic identification of wheezes in breathing. It has been demonstrated that certain individual features (MFCC, tonality index) are much more accurate in detection of wheezes. However their computation requirements are higher than those of simpler time-domain features. In addition, it has also been shown that while the use of multiple features does increase the classification accuracy in some cases, the gain in performance becomes very limited after a certain number of features. While the classifier used in this work is very simple, the use of other more complex classifiers such as support vector machines, artificial neural networks, etc. may help to increase the classification performance at the added cost of computational complexity. Thus, it is important to take all the competing requirements into account when selecting a feature for wheeze detection in different applications. The results presented in this paper will provide highly useful insights to address these requirements for the development of wheeze detection algorithms.