Automatic Classification of Healthy/TB Chest X-ray using DeepLearning
K. Vijayakumar, Mohammad Nazmul Hasan Maziz, Swaetha Ramadasan, S. Prabha, K. Sri Nirmal Kumaar
Abstract
Lung is a vital internal organ responsible for the respiration process. Abnormality in the lung causes mild to severe illness and may lead to death. Tuberculosis (TB) is a common disease in people, and it is very important to find and treat it quickly. At the clinical level, chest x-rays are used to diagnose TB, and based on the results, the right treatment needs to be started. The suggested study aims to use DenseNet (DN) variants and deep transfer learning (DTL) to sort chest X-rays into two groups: healthy and TB. The plan that was put into action has five steps: collecting and resizing data, deep-features mining using DN variants, feature reduction and serial features fusion, and binary classification and verification. In this paper, the performance of the system that was built is tested using both separate and combined features. Several binary classifiers are used to complete the classification job. The results of this study show that the detection accuracy is >91% with individual features and 99% when fused features are taken into account. This proves that the plan that was put in place works better on the chosen image database.