Computerized analysis of respiratory sounds during COPD exacerbations
Introduction
Chronic obstructive pulmonary disease (COPD) is a major cause of mortality and morbidity and an important factor accounting for rising healthcare costs worldwide [1]. The overall prevalence of COPD in adults is higher than 10% of population [2], and it is estimated to be the fourth leading cause of death worldwide in 2030 [3]. This disorder is caused by long-term exposure to noxious particles and gases and characterized by a non-fully reversible chronic airflow limitation. Chronic inflammation leads to alterations such as small-airway obstruction, remodelling and parenchymal destruction. COPD leads to markedly reduced exercise capacity and, eventually, disability.
Acute exacerbation of COPD (AECOPD) is a major event in the natural course of the disease characterized by a worsening of respiratory symptoms. AECOPD is associated with significant mortality, adversely affects patient's quality of life and represents a huge socioeconomic burden mainly due to hospitalizations [4]. Key symptoms of exacerbations are increased breathlessness, cough and sputum production that is beyond normal day-to-day variations [1]. Normally, exacerbations appear about twice or three times per year, depending on COPD severity, causing a worsening in the patient's health status [4]. Diagnosis of AECOPD is made clinically when the patient presents acute worsening of dyspnea, cough or sputum production. Although respiratory sounds such as wheezes are reported by patient's self-assessment in 35% of AECOPD [5], little attention has been given to such a sign.
The course of exacerbations is characterized by a significant increase in airway obstruction and abnormal bronchial mucus production [6]. Respiratory sounds such as wheezes and rhonchi are a manifestation of airway narrowing and mucus secretion, respectively [7], [8], and are hallmark symptoms related to the pathophysiology of AECOPD. The changes to the pulmonary structure induced by an exacerbation affect sound transmission to the surface of the chest wall [9]. Therefore, changes in respiratory sounds are one of the clinical signs reported in exacerbation episodes. It is generally accepted that normal vesicular lung sounds are usually replaced by adventitious sounds like wheezes or rhonchi in AECOPD. Monitoring changes in respiratory sounds before, during and after a patient-specific therapy is seen as a valuable clinical strategy on which physicians can rely to confirm clinical improvement in patients.
Over the past decades, computational methods used for recording and analyzing respiratory sounds have overcome the many limitations of conventional auscultation. But despite being one of the most distinguishing characteristics of episodes of exacerbation, the changes and peculiarities of respiratory sounds during an exacerbation have been barely studied. Computerized analysis of relative differences in respiratory sound intensity has been recently reported to be potentially useful in distinguishing acute dyspnea caused by congestive heart failure (CHF), COPD or asthma during acute exacerbations [10]. To our knowledge, there are no significant studies that have examined the evolution of respiratory sounds in patients hospitalized for AECOPD during the period of hospital stay in order to generate clinical and diagnostic applications.
The aim of this work is to examine respiratory sounds in patients with COPD hospitalized for AECOPD and find substantial differences in the sounds evolutionary behaviour that may allow the establishment of exacerbation typologies with possible clinical implications. Therefore, the purpose of this study is to provide an answer to the hypothesis that a computerized system can detect changes in respiratory sounds during COPD exacerbations and that these changes are consistent with those assessed by physicians, thereby aiding in the detection of AECOPD.
Section snippets
Subjects
A group of patients hospitalized due to AECOPD in the Pulmonology and Allergy Unit of the University Hospital Puerta del Mar of Cádiz (Spain) were followed during their hospital stay. An intentional, non-probabilistic sample of 53 individuals was studied. Because the recruited participants had mild to very severe airflow limitation, they represented a broad spectrum of COPD patients. Respiratory sounds were recorded from admission to discharge. The hospital's research ethics committee approved
Sounds recording
Sounds were recorded with an electret microphone with coupling chamber and flat response between 50 and 18.000 Hz. Electronics was embedded in a special housing tailored for self-use (Fig. 1). Sensor was placed over the trachea on the suprasternal notch and handled by the patients themselves. Sampling rate was 8000 Hz. Analog to digital conversion was performed with 16-bit quantization. Both clinical history and sound files were assigned to a codified electronic patient record, especially
Dimensionality reduction and exploratory data analysis
Explorative data analysis techniques and unsupervised clustering methods were applied in order to study the resulting dataset built from the 53 COPD patients. Due to the difficulty in understanding the underlying information of such a complex data model, and to learn about its multidimensional structure and the potential existence of subsets of patients with observable differences in clinical practice, the interrelationships among the attributes were evaluated by applying a data-driven
Results
Patients' morphological, smoking and functional characteristics are presented in Table 2. Application of PCA to the dataset led to the conclusion that the three first components accounted for over 90% of the sample variance, and the first two for more than 83%. In order to perform a better discriminating visual exploratory analysis, the first two components were considered for further processing stages.
Biplots show inter-unit distances, variances and correlations of variables of datasets. They
Discussion
The main finding of this study is that two distinct groups or clusters of patients with AECOPD arose from the unsupervised analysis of parameters derived from the respiratory sounds processing.
While conventional auscultation with a stethoscope is subjective and hardly comparable [27], acoustic studies with computer systems could provide objective aid, with better sensitivity and reproducibility of the results. Respiratory acoustic analysis allows quantifying changes in the respiratory sounds,
Conclusions
This study suggests that COPD population does not show a homogeneous spectrum but two different respiratory sounds patterns that can be differentiated in the course of AECOPD through computerized analysis. To our knowledge, the aforementioned finding has not been described in other previous works, and could enable the detection or follow-up of exacerbations through remote monitoring of respiratory sounds. Furthermore, the results of this study open the door to further studies that can assess
Summary
In computerized analysis of respiratory sounds, most efforts have focused on the improvement of automatic detection systems for certain adventitious sounds. However, this strategy has some weak points, such as inter- and intra-subject variability. Some exacerbated patients do not present relevant respiratory adventitious sounds.
There have recently been several attempts to define COPD exacerbation phenotypes. The attempts are being mainly based on biological markers. These phenotypes could
Authors’ contributions
The work presented here was carried out in collaboration between all authors: they all defined the research theme, read abstracts, selected reviews for full text assessment, reviewed related papers, designed, conducted the study, interpreted its results, revised the manuscript and approved the final version. DSM managed the review process. DSM and AL drafted the full article and revised it critically for intellectual content. DSM carried out the computerized processing of signals, and SA and
Conflict of interest statement
The authors declare no conflict of interest.
Acknowledgements
This work was supported in part by the Ambient Assisted Living (AAL) E.U. Joint Programme, by grants from Ministerio de Educación y Ciencia (Ministry of Education and Science) of Spain and Instituto de Salud Carlos III under Projects PI08/90946 and PI08/90947.
Daniel Sánchez Morillo received his Engineering degree in Telecommunications from the University of Seville, Spain and his Ph.D. degree from the University of Cádiz, Spain. He is Professor at the School of Engineering of Cádiz and currently is with the Biomedical Engineering and Telemedicine Researching Group. His research interests are in biomedical signal processing, e-health, HMI and ambient assisted living. He is a member of the Spanish Biomedical Engineering Society.
References (46)
- et al.
Pulmonary auscultation in the era of evidence-based medicine
Rev. Mal. Resp.
(2008) - et al.
Tracheal and lung sounds repeatability in normal adults
Resp. Med.
(2003) - et al.
Explorative data analysis techniques and unsupervised clustering methods to support clinical assessment of chronic obstructive pulmonary disease (COPD) phenotypes
J. Biomed. Inf.
(2009) - et al.
FCM: the fuzzy c-means clustering algorithm
Comput. Geosci.
(1984) - et al.
Computerized lung sound analysis as diagnostic aid for the detection of abnormal lung sounds: a systematic review and meta-analysis
Resp. Med.
(2011) - et al.
Wheeze detection based on time-frequency analysis of breath sounds
Comput. Biol. Med.
(2007) - et al.
Adenosine bronchial provocation with computerized wheeze detection in young infants with prolonged cough: correlation with long-term follow-up
Chest
(2004) Pattern recognition methods applied to respiratory sounds classification into normal and wheeze classes
Comput. Biol. Med.
(2009)- et al.
Pneumonic vs nonpneumonic acute exacerbations of COPD
Chest
(2002) - et al.
Identification of a predominant COPD phenotype in clinical practice
Resp. Med.
(2008)
BOLD Collaborative Research Group. International variation in the prevalence of COPD (the BOLD study): a population-based prevalence study
Lancet
Projections of global mortality and burden of disease from 2002 to 2030
PLoS Med.
Acute exacerbations of COPD in the United States: inpatient burden and predictors of costs and mortality
COPD
Time course and recovery of exacerbations in patients with chronic obstructive pulmonary disease
Am. J. Resp. Crit. Care. Med.
Structural and functional co-conspirators in chronic obstructive pulmonary disease exacerbations
Proc. Am. Thorac. Soc.
Auscultation in the diagnosis of respiratory disease in the 21st century
Postgrad. Med. J.
Respiratory sounds. Advances beyond the stethoscope
Am. J. Resp. Crit. Care. Med.
Lung sound patterns help to distinguish congestive heart failure, chronic obstructive pulmonary disease, and asthma exacerbations
Acad. Emerg. Med.
Correlation of constant flow rate with frequency spectrum of respiratory sounds when measured at the trachea
IEEE Trans. Biomed. Eng.
Bronchial provocation determined by breath sounds compared with lung function
Arch. Dis. Child.
Automated spectral characterization of wheezing in asthmatic children
IEEE Trans. Biomed. Eng.
A model of acoustic transmission in the respiratory system
IEEE Trans. Biomed. Eng.
Cited by (40)
Comparative study of respiratory sounds classification methods based on cepstral analysis and artificial neural networks
2024, Computers in Biology and MedicineNonlinear time series and principal component analyses: Potential diagnostic tools for COVID-19 auscultation
2020, Chaos, Solitons and FractalsCitation Excerpt :The present work attempts a detailed analysis of VB and BB sound signals through spectral, nonlinear time series and complexity analysis to extend to COVID 19 auscultation. Principle component analysis is also employed to separate the VB and BB sound signals for computer-aided automated detection [27]. Nowadays, digitization and automation of the medical field employing computer-aided techniques have made the diagnosis accurate and quick.
Enhancing our understanding of computerised adventitious respiratory sounds in different COPD phases and healthy people
2018, Respiratory MedicineCitation Excerpt :As respiratory sounds are directly related to the movement of air within the tracheobronchial tree [9], the changes in respiratory mechanics related with AECOPD may be primarily detected by changes in respiratory sounds, namely adventitious respiratory sounds (ARS, crackles and wheezes). Recent studies have shown respiratory sounds ability to differentiate between groups of patients with stable and exacerbated COPD [10] and to characterise AECOPD into two phenotypes, based on computerised analysis [11]. Nevertheless, there is little information available on the time course of respiratory sounds changes during recovery from AECOPD, within the same group of patients.
Unwrapping aortic valve dysfunction through complex network analysis: A biophysics approach
2022, Journal of Applied Physics
Daniel Sánchez Morillo received his Engineering degree in Telecommunications from the University of Seville, Spain and his Ph.D. degree from the University of Cádiz, Spain. He is Professor at the School of Engineering of Cádiz and currently is with the Biomedical Engineering and Telemedicine Researching Group. His research interests are in biomedical signal processing, e-health, HMI and ambient assisted living. He is a member of the Spanish Biomedical Engineering Society.
Sonia Astorga received her Medical Degree from the Faculty of Medicine (University of Cádiz, Spain). She has been a scholarship holder in the Pulmonology and Allergy Unit of the Puerta del Mar University Hospital of Cádiz, where she is researching towards the Ph.D. degree. Her scientific interest is focused to the areas related to allergy and pulmonology.
Miguel Angel Fernandez is Professor at the School of Engineering of Cádiz, Spain, from which he received his Bachelor in Engineering. He currently lectures in Automatics and Systems Engineering in the University of Cádiz and he is with the Biomedical Engineering and Telemedicine Researching Group where he is researching towards the Ph.D. degree.
Antonio León is Dr. in Medicine and Associate Professor at the Faculty of Medicine (University of Cádiz, Spain). He is currently the Head of the Pulmonology and Allergy Unit of the Puerta del Mar University Hospital of Cádiz. He coordinates the COPD care management process by the Ministry of Health (Regional Government of Andalusia) and is a member of the Working Group of the COPD South Association of Pulmonologists and of the Spanish Society of Respiratory Pathology (SEPAR).