CLUSTERING OF LUNG DISEASE BASED ON CHEST X-RAY USING DIMENSIONS FRACTAL BOX COUNTING AND K-MEDOIDS

Abstract

Lungs are one of the respiratory organs in humans that are susceptible to disease because they are easily infected. The respiratory tract is an internal human organ that is attacked by the SARS-Cov-2 virus that causes the corona disease (Covid-19). In addition, Lung Opacity and Viral Pneumonia are also diseases that occur in the lungs. Diseases of the lungs can be detected using chest x-ray images, but to diagnose from the results of the chest x-ray image it is not easy to analyze, in addition to the need for a fairly good x-ray result expert knowledge is also needed to determine the type of disease. The mathematical method that studies irregular shapes is fractal geometry and fractal dimensions have been widely applied in the medical field in determining the type of disease. In this research, lung deseases will be classified based on chest x-ray images. There are 100 chest x-ray images that will be processed using segmentation that result a region of the lungs. The region is used to determine the spots of the type of lung disease using Canny edge detection. Then by using fractal box counting, the fractal dimension value will be calculated for use in clustering. The results of the experiment using the K-Medoids clustering method with four clusters, that is Covid-19, Lung Opacity, Normal Lung, and Viral Pneumonia have an accuracy of 87%.