Deep Learning-Driven Tuberculosis Screening Via Optimized Densenet on Chest Radiographic Data

A. Lohithasai1, K. Hema Kumar2, Kalagatla Sai Divya3 and Eriki Ananda Kumar4*

1, 2, 3, 4 Department of Computer Science and Engineering, SV College of Engineering, Puttur, AP, India4*erikiananda@gmail.com

Abstract— In this paper, we describe an improved DenseNet deep neural network model designed specifically for exploiting chest X-ray images to detect tuberculosis (TB). Utilizing the intrinsic benefits of DenseNet's densely interconnected layers, we suggest innovative architectural adjustments and optimization techniques to enhance the model's effectiveness and efficiency. We show notable improvements over current approaches in terms of sensitivity, specificity, and overall accuracy through thorough testing on a wide range of datasets. Our enhanced DenseNet model shows potential for improving diagnostic capabilities in clinical settings by robustly recognizing TB symptoms in chest X-ray images."Tuberculosis (TB) is a highly contagious and sometimes lethal infectious illness that affects millions of people worldwide. Early identification of tuberculosis is essential for prompt treatment and containment of the disease's progress. This study suggests a unique deep learning model called CBAMWDnet for the detection of tuberculosis (TB) in images from chest X-rays (CXRs). The cornerstone of the model is the Convolutional Block Attention Module (CBAM) and Wide Dense Net (WDnet) architecture, which was developed to effectively collect spatial and contextual information in the images. The performance of the proposed model is evaluated on a large dataset of chest X-ray images and compared with many state-of-the-art models.Keywords: TB, DL, DCN, Chest X-ray