Extraction in Detecting Tuberculosis X-Ray Results using Histogram of Oriented Gradients

Abstract

Image processing is a very popular study in research. Image processing research developed from starting to improve images that have lost pixels or adding pixels to make the image clearer in providing and conveying information. However, with the support of various image processing computational techniques, it is not only conveying information but also applying machine learning techniques to carry out lessons from the available data. In this study, a training was implemented in an X-ray image of Tuberculosis. Tuberculosis is made due to the presence of bacteria that have gathered so that the results of X-rays of Tuberculosis are processed by extracting features to detect whether the results of the Tuberculosis X-ray image are positive or negative. The very simple feature extract is processed to detect the Histogram of Oriented Gradients (HOG) because it applies the gradient function to be mathematically processed in detecting Tuberculosis. In this paper, the results using HOG feature extraction based on the percentage of positive tuberculosis are 70.90%, the proportion of negative diagnosis results is 29.10%, the proportion of negative diagnosis results is 72.72%, and the positive diagnosis results are 27.28%.