Classification of Lung Tuberculosis using Non Parametric and Deep Neural Network Techniques

Intelligent systems are emerging rapidly to mimic the learning capability of human brain. These systems are developed to learn, adapt, act and make decisions autonomously based on the problem, commonly named as Computer Aided Disease Diagnosis (CAD). In this research paper, a CAD system is developed and analyzed for Tuberculosis (TB) disease types using non parametric (Support Vector Machine) and deep neural network (Convolutional Neural Network) learning techniques. The proposed algorithms are evaluated for 3D Computed Tomography (CT) images of ImageCLEF 2018 Tuberculosis using appropriate quantitative metrics. From the results, it has been inferred that CNN has resulted 7% improved accuracy than kernel based SVM classification.