An Automated System for Classification of Chronic Obstructive Pulmonary Disease and Pneumonia Patients Using Lung Sound Analysis

Abstract

Chronic obstructive pulmonary disease (COPD) and pneumonia are two of the few fatal lung diseases which share common adventitious lung sounds. Diagnosing the disease from lung sound analysis to design a noninvasive technique for telemedicine is a challenging task. A novel framework is presented to perform a diagnosis of COPD and Pneumonia via application of the signal processing and machine learning approach. This model will help the pulmonologist to accurately detect disease A and B. COPD, normal and pneumonia lung sound (LS) data from the ICBHI respiratory database is used in this research. The performance analysis is evidence of the improved performance of the quadratic discriminate classifier with an accuracy of 99.70% on selected fused features after experimentation. The fusion of time domain, cepstral, and spectral features are employed. Feature selection for fusion is performed through the back-elimination method whereas empirical mode decomposition (EMD) and discrete wavelet transform (DWT)-based techniques are used to denoise and segment the pulmonic signal. Class imbalance is catered with the implementation of the adaptive synthetic (ADASYN) sampling technique.

Lung disease recognition methods using audio-based analysis with machine learning

“…Most existing projects for lung disease diagnosis using respiratory sounds follow a three-stage machine learning pipeline: 1) the preprocessing, which involves removing unwanted noise and preparing the sound data for further analysis using audio filtering and noise-reduction techniques, 2) the feature extraction, which involves extracting relevant characteristics from the preprocessed sound data using signal processing methods like spectral analysis [ [39] , [40] , [41] , [42] ], cepstral analysis [ [43] , [44] , [45] ], wavelet transforms [ [46] , [47] , [48] ], and statistical analysis [ 49 ], 3) the classification, which uses extracted features to categorize the sounds as belonging to different disease categories. Popular classifiers include K-nearest Neighbors [ [50] , [51] , [52] , [53] , [54] ], Support Vector Machines [ [55] , [56] , [57] , [58] , [59] ], Gaussian Mixture models [ 60 , 61 ], and Artificial Neural Networks [ 36 , 55 ].…”

“… Respiratory Sounds Dataset ICBHI (RSD) 2017 [ 69 ] Contains normal audio signals and three types of adventitious sounds: wheezes, crackles, and combined wheezes/crackles. [ 30 , 48 , [63] , [64] , [65] , [66] , [67] , [68] ] HF_Lung_V1 [ 21 ] Includes 9765 lung sound audio files (15 s each), labels for exhalation, inhalation, and irregular and regular adventitious sounds (rhonchus, wheeze, stridor). [ 21 ] Respiratory-Database@TR [ 71 ] Short recordings, 12-channel lung sounds for each patient, multi-channel analysis, 5 COPD severity levels.…”

“…Annotations include the start/end timings of each respiratory cycle and the presence of wheeze/crackle. [ 30 , 48 , 54 , 62 , 72 , 79 ] …”

Exploring machine learning for audio-based respiratory condition screening: A concise review of databases, methods, and open issues

“…Similarly, the SVM classifier also showed an accuracy of around 99% in distinguishing COPD, pneumonia, and health subjects based on the International Conference on Biomedical and Health Informatics (ICBHI) 2017 database. 50 In 2022, researchers further verified the promise that ML can be used to identify COPD subjects from healthy controls in a private but clinically validated voice data set, and according to their study, Compare2016 feature set developed by openSMILE toolkit presented better accuracy than other features. 77 Based on the public ICBHI data base, Monaco et al 76 compared the performance of RF, MLP, and SVM by exploring 33 acoustic features with their statistics, although MLP yielded the highest accuracy of 85%, the performance difference from other models is marginal: their accuracy ranged from 81% to 85%.…”

Review on Pneumonia Image Detection: A Machine Learning Approach

“…According to the report released during World Pneumonia Day, it is estimated that more than 11 million infant children below the age of 5 years are likely to die from pneumonia by the year 2030 [5]. In the early nineteenth century, pneumonia was considered one of the significant causes of death amongst people.…”