CNN based Framework for intelligent Diagnosis of Tuberculosis using Chest Radiographs

Abstract

Medical Image Processing plays a major role in optimized identification of various diseases. In many parts of the world, tuberculosis is a serious health problem. Even in today's environment, diagnosing tuberculosis (TB) is difficult. The mortality role of those affected with TB is high due to the undiagnosed and untreated nature. Early detection of tuberculosis (TB) using X-rays of the lungs and classification to assist the treatments needed to improve their day-to-day routines. Early identification of the TB the lung X rays are segmented using Particle Swarm Optimization scheme. Features are extracted from the segmented lung Region of Interest using the texture and the shape features. Prominent Features are identified using a genetic algorithm. The reduced set of features are classified using neural network thus enabling the images to be classified as Normal or Abnormal. The accuracy, recall and, sensitivity achieved by the methodology have been reported in this paper.