Improving Life-threatening Lung Diseases Classification using Hybrid SMOTE-ENN with assorted Machine Learning Classifiers | ||
International Journal of Telecommunications | ||
Volume 04, Issue 02, July 2024, Pages 1-17 PDF (2.04 M) | ||
Document Type: Original Article | ||
DOI: 10.21608/ijt.2024.293555.1056 | ||
Authors | ||
Mostafa albanhawy* 1; Abeer Khalil2; Hossam Moustafa3 | ||
1Department of Electronics and Communications, Mansoura University, Daqahliyah , Egypt | ||
2Electronics and communications department, Faculty of Engineering, Mansoura University, Egypt | ||
3Department of Electronics and Communications Engineering at the Faculty of Engineering, Mansoura Uni-versity | ||
Abstract | ||
Chest radiography is one of the most common diagnostic tools for diagnosing and managing bronchopneumonia and other lung diseases. In this paper, a classification strategy was proposed for identifying infection in Chest X-ray images. We collected 7545 x-ray chest images from an openly accessible X-ray database and separated them into three classes: healthy individuals, persons suffering from pneumonia, and additional COVID-19 patients. The contrast limited adaptive histogram equalization (CLAHE) method was used to improve the quality of the X-ray images. The oriented gradient histogram (HOG) is used. The classification of datasets in medi-cine sometimes is hindered by the problem of having unequal datasets. In the solving of this problem, which occurs during imbalanced data classification in medical diagnosis, we introduce a hybrid sampling technique called SMOTE-ENN that is a combination of the Synthetic minority oversampling technique (SMOTE) and Edited Nearest Neighbors (ENN). The support vector machine (SVM), k-Nearest Neighbors (k-NN), and Random Forest Classifier (RFC) used to clas-sify the images, with classification rates of 99.47%, 98.70%, and 98.47%, respectively, on a test dataset of 1504 images. These findings may help to detect COVID-19 and pneumonia diseases more effectively. | ||
Keywords | ||
SVM; CXR; Oriented Gradients Histogram; Synthetic minority over-sampling technique (SMOTE); Edited Nearest Neighbor (ENN) | ||
Statistics Article View: 273 PDF Download: 168 |