Klasifikasi Anomali pada Gambar Rontgen Dada dengan Metode Machine Learning Menggunakan Histogram of Oriented Gradient dan Random Forest

Abstract

Before confirming that there is an abnormal condition, doctors use radiological examinations as the first step in diagnosing lung disease. To classify unusual conditions, an examination is carried out via a chest x-ray. Therefore, this study aims to build a model for classifying abnormal objects using the Histogram of Oriented Gradient (HOG) and Random Forest (RF) algorithms on chest x-ray images, where there will be 12 abnormal objects to be classified. Among them are: Atelectasis, Cardiomegaly, Concolidation, Infiltration, Nodule, Mass, Emphysema, Fibrosis, Pleural Effusion, Pneumothorax, Pneumonia and No_Finding. There are several techniques used to improve model accuracy when building a training model, namely sqrt and log2. The best accuracy results have been obtained from the Random Forest model by training using an n-estimator of 100 and the max features sqrt is 92%, with these results it can be concluded that the Histogram of Oriented Gradient and Random forest methods used in this study can classify X-ray results properly. In addition to building the model, the authors also developed a desktop-based application that aims to assist general practitioners in facilitating the process of diagnosing disease by analyzing lung results. This application uses image processing technology to classify signs of disease seen on X-ray images of the lungs. This desktop-based application is expected to help doctors or related health workers to simplify the process of disease diagnosis and can be utilized as a good and useful learning tool.