Ensembling Intelligent Models to Design an Efficient System for Prediction of Pulmonary Tuberculosis

Tuberculosis, as a contagious pulmonary disorder, may spread easily from an infected person to a non-infected person. The Microbiologically Confirmed Pulmonary TB Survey done from 2019–2021 in a population aged ≥ 15 years shows Delhi with 534 per lakh at the top among the high prevalence 20 states of India. The gravity of incidences and the increasing mortality rate in tuberculosis indicate limited progress in technological advancements in AI-based diagnosis systems. The prediction of pulmonary tuberculosis through preprocessing of image inputs is proposed here with reliable intelligent techniques. DL needs huge training on high-quality data samples. Normally, chest X-rays have low contrast. So three image enhancement methods are applied in image preprocessing: UM, HEF, and CLAHE. An ensembling model using deep learning networks and machine learning algorithms may reduce heavy computational work. The pivotal features gained from the deep neural networks will then be grouped and processed into the classifiers. The machine learning algorithms will predict positive and negative pulmonary tuberculosis cases. The proposed model will undergo n-fold cross- validation, and furthermore, its accuracy will be evaluated.