Multi Res U-Net Based Image Segmentation of Pulmonary Tuberculosis Using CT Images

The Pulmonary tuberculosis (TB) is a contagious, infectious disease that attacks the lungs. The process of separating the TB affected region from the normal region of the lung is called as segmentation. To acquire the process of segmentation the CT image of the lung is used. The process produces an accurate result as an outcome at the initial stage of the pulmonary tuberculosis. This helps the medical practitioner to proceeds the decision faster for providing a needed treatment. There are two methods for diagnosing the pulmonary tuberculosis and they are manual recognizing and computer aided recognizing. In the manual recognizing it requires an expert and also for the expert it is an unenviable and long-lasting task. Most of the computer aided recognizing method uses traditional way of image segmentation. A new method is proposed in this paper which is known as is known as MR U-Net image segmentation model for segmenting the pulmonary tuberculosis affected region. Thus the system consist of the deep learning concept, which helps to train the system with only few datasets and it is also used to test `n' number of datasets. The detection accuracy and the system accuracy are the two parameters that are used to evaluated the performance of the system.