Demystifying Computational Techniques Used to Diagnose Tuberculosis

Abstract

Tuberculosis is one of the most deathful diseases in the entire world and remains a major reason for death worldwide. In 2017, around 10.0 million people infected with tuberculosis. For detecting disease in the medical field using the computational technique, MRCNN & UNET Model has achieved impressive accuracy across multiple datasets like brain tumor (Multimodal Brain Tumor Image Segmentation (BRATS 2015) datasets), glaucoma and other based on data collected. This Research is based on the detection of WBC (White Blood Cell) form the stained Microscopy image, by collecting past data form patients. In India, medical patient data is not stored anywhere systematically, we made extra effort to find hidden patterns from data. The article deeply discusses the various approaches to diagnose tuberculosis. It summarizes the advantages and disadvantages of the existing techniques and why deep learning technology use in the various medical diagnosis process. In this study, we propose a fully automatic method for WBC segmentation, which is developed using MRCNN based deep convolutional networks. Proposed technique was evaluated on a dataset that contains 2500 stained microscopy images are used to train the system and 540 images are used to test the system. We have achieved nearly 92% accuracy with a low false-positive type error.