Tuberculosis: Diagnosis by Image Processing

Abstract

Tuberculosis is one of the first human diseases of which there is evidence, it´s very harmful and easy to spread through the air. One way to detect tuberculosis is by chest x-rays, by analyzing the x-ray you can obtain the detection of any abnormality (parenchymal, ganglionic or pleural). This paper presents a method that allows the presence of tuberculosis in medical X-ray images to be identified. Three methods of classification were implemented for the evaluation of the method: Support Vector Machine, Logistic Regression and K-Neighbors Classifier. Two classification scenarios were implemented: cross validation and training and test sets. The results obtained allow us to see the viability of the proposed method.