DeepXray: A Deep Learning Based system for Tuberculosis Detection and Severity Prediction in Chest X-rays
R Dhruthi, Bandi Monica Sreevalli, Sachi Kulkarni, Surabhi Gudla, B J Sandesh
Abstract
Tuberculosis (TB), a highly contagious bacterial illness, poses a threat to global health due to the difficulty in diagnosing it quickly and accurately. The DeepXray project tackles the problem with four deep learning models by using the Montgomery and Shenzhen datasets. These algorithms demonstrate promising results for automating TB detection and severity prediction using chest X-rays. The deep learning models used were VGG-16, DenseNet-121, Resnet-50, and Alexnet; the corresponding accuracies were 86.25, 89.25, 84.16, and 80.1%. The experiment demonstrates astounding precision and holds the potential to completely transform the efficacy of diagnostics, especially in environments with little resources where traditional procedures are ineffective.