Segmentation Techniques for Detection of Tuberculosis Using Deep Learning: A Review

Among many uses of image segmentation are scene understanding, image analysis for medical purposes, robotic understanding, video monitoring, augmented reality (AR), and compression of images. There are many different image segmentation computational methods available. Due to growing popularity of models using deep learning in vision applications, many current efforts have focused on advancing deep learning-based methods for image segmentation. This survey provides a state-of-the-art overview of state-of-the-art in field of segmentation regarding Tuberculosis (TB), including a wide range of ground-breaking computational models such as Unet, Vnet, SegNet, and Fully Convolutional Neural Networks (FCN). X-rays of the lungs are the primary method used by medical professionals in previous research for diagnosing tuberculosis (TB). Mycobacterium tuberculosis has been infectious agent that causes tuberculosis (TB), condition that can spread from person to person. Around the world in year 2019, tuberculosis was responsible for mortality of almost 1.4 million people. Using deep learning algorithms for classification has improved TB detection accuracy nearly comparable with that of a human doctor. When applied to lung segments rather than the full X-ray, techniques for classification improve the likelihood of detecting tuberculosis.