Machine Learning-Based Classification System for Tuberculosis Detection Using Locally Collected Radiographs

Le résumé fourni par la source

Abstract: Tuberculosis (TB) is still a severe public health problem. Pakistan also has a high burden of TB and is on number 6th among in top 30 countries nationwide. There are many Computer-aided detection (CAD) systems to detect TB. But the problem is that all the available systems are developed and tested within developed states and when they are tested for middle-income countries their performance varies. The main purpose of this study is to develop a machine learning-based classification system for Tuberculosis detection using locally collected radiographs which is specifically designed for the middle-income countries according to their socio-economic factors. In this study, we first collect a dataset comprising of X-ray images, The Digital Imaging and Communications in Medicine (DICOM) file format. These X-ray images were collected from the Provincial TB Control Program in Punjab, which had conducted Chest camps using mobile X-ray vans in remote areas across various districts within the Punjab Province. We apply data augmentation and a convolutional neural network (CNN) from the beginning. that the term “Conv” shows the convolution layer with the 380 TB positive and 421 Normal X-rays. In training accuracy 99.53%, validation accuracy 100%, Test accuracy 99.17%, Test loss 0.0844.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.