Lung sounds can vary according to various respiratory diseases of the person. Specialist physicians use these sound data to make a diagnosis. Diagnostic success varies according to the physician’s experience. computer-aided diagnostic systems can help physicians in this regard. In this study, disease diagnosis system was developed by using lung sound data obtained by auscultation method. In experimental studies, various machine learning methods have been tried on 20 normal, 20 ral and 20 rhoncus sound data taken from 60 patients. In addition, the data set was tripled with two different artificial data generation methods. The results obtained by applying k- Nearest Neighbor (kNN), Support Vector Machine (SVM), Naive Bayes, Decision Tree and Random Forest Classifier to all data obtained by real data set and artificial data production are presented. A 95% accuracy value was obtained with 10 cross- validation using the Naive Bayes classification method. In the results obtained after artificial data production, an accuracy value of 94% was obtained with 10 cross-validation with the kNN method.
Paper
Full text
Classification of Abnormal Respiratory Sounds Using Machine Learning Techniques
Semantic Scholar · Medicine · 2020
Abstract
Lung sounds can vary according to various respiratory diseases of the person. Specialist physicians use these sound data to make a diagnosis. Diagnostic success varies according to the physician’s experience. computer-aided diagnostic systems can help physicians in this regard. In this study, disease diagnosis system was developed by using lung sound data obtained by auscultation method. In experimental studies, various machine learning methods have been tried on 20 normal, 20 ral and 20 rhoncus sound data taken from 60 patients. In addition, the data set was tripled with two different artificial data generation methods. The results obtained by applying k- Nearest Neighbor (kNN), Support Vector Machine (SVM), Naive Bayes, Decision Tree and Random Forest Classifier to all data obtained by real data set and artificial data production are presented. A 95% accuracy value was obtained with 10 cross- validation using the Naive Bayes classification method. In the results obtained after artificial data production, an accuracy value of 94% was obtained with 10 cross-validation with the kNN method.