Detection of Tuberculosis Bacilli Using Deep Learning

A harmful bacteria known as MT is the primary cause of the communicable disease tuberculosis (TB) (Mycobacterium Tuberculosis). This disease can be treated but because of its contagious nature, it comes under the category of the world's scariest as well as the deadliest disease. From this point of view, the detection of Tuberculosis (TB) at a very early stage is mandatory to provide successful treatment and then prevent it from spreading to other people also. Due to their efficiency and inexpensive cost, microscopy tests using ZN-stained images are typically favored in underdeveloped nations. However, manual detection of TB bacilli under a microscope is a very time-consuming and tedious task. In the proposed work patch wise detection strategy using CNN is done for which a set of image patches are extracted from the supplied list of annotated images. The patches were then categorized into negative and positive patches. Data augmentation is used to balance out the quantity of positive and negative patches.