Detection of Tuberculosis Disease with Convolutional Neural Networks

Abstract

Tuberculosis is an infectious disease caused by bacteria called Mycobacterium tuberculosis. Tuberculosis is still an important public health problem worldwide and is common especially in developing countries. This respiratory disease can cause serious symptoms, especially affecting the lungs. Symptoms of tuberculosis include prolonged coughing, shortness of breath, chest pain, weakness, fever, night sweats, and malaise. The diagnosis of the disease is made by clinical signs as well as biomedical imaging methods and laboratory tests. These imaging modalities include techniques such as x-rays, computed tomography (CT), and magnetic resonance imaging (MRI). Early diagnosis of tuberculosis disease is of great importance in terms of treatment and prevention of the spread of the disease. The use of deep learning methods to classify biomedical images of tuberculosis disease can accelerate the diagnosis process, increase accuracy and guide treatment more effectively. In this study, it aims to be an important step in the classification of tuberculosis disease with deep learning. The generated CNN network, parameter values, layers used, complexity matrices obtained for verification data, accuracy and loss graphs are shown in detail. In our study, the success rate was increased by using a different network structure than the neural networks used in the literature. Approximately 98% success was achieved with the proposed CNN model.