Automated Image-Based Tuberculosis Diagnosis Using 2D Convolutional Neural Networks

Abstract: Tuberculosis chest radiography image datasets are used to train convolutional neural networks designed for automated diagnosis of tuberculosis. First, a convolutional neural network of suitable complexity is designed, trained, tested and validated on the tuberculosis chest radiography image sequences. The resulting artificial intelligence models could then be refined and packaged into modules for the automated detection of tuberculosis in chest radiography images and could form part of a comprehensive artificial intelligence-driven framework for the detection, prediction, diagnosis and management of a wide variety of health conditions that could play a crucial clinical decision support role.

Keywords: Artificial Intelligence (AI), Convolutional Neural Network (CNN), TensorFlow, Healthcare System, Automated Disease Diagnosis and Prediction, Tuberculosis.