Machine Vision Approach for Diagnosing Tuberculosis (TB) Based on Computerized Tomography (CT) Scan Images

Tuberculosis is curable, still the world’s second inflectional murderous disease, and ranked 13th (in 2020) by World Health Organization on list of leading death causes. One reasons for its fatality unavailability modern technology human experts early detection. This study represents a precise reliable machine vision-based approach detection in lung through Symmetry CT scan images. TB spreads irregularly, which means it might not affect both lungs equally, only some part lung. That’s why regions interest (ROI’s) from infected normal images were selected after pre-processing i.e., selection/cropping, grayscale image conversion, filtration, Statistical texture features extracted, 30 optimized using F (Fisher) + PA (probability error average correlation) MI (mutual information) final optimization 6 most selected. Several supervised learning classifiers used to classify between Artificial Neural Network (ANN: n class) based classifier Multi-Layer Perceptron (MLP) showed comparatively better probably best accuracy 99% with execution time less than second, followed Random Forest 98.83%, J48 98.67%, Log Boost 98%, AdaBoostM1 97.16% Bayes Net 96.83%.