Automated Pulmonary Tuberculosis Screening Based on Deep Convolutional Neural Networks Using Chest Radiographs

Abstract

Pulmonary Tuberculosis (PTB) is considered one of the most dangerous illnesses if left untreated. Researchers and physicians are taking more interest in automated screening of pulmonary tuberculosis using chest radiographs. The computer-aided diagnosis (CAD) system for TB is one of the automated methods for early detection and treatment. Previous literature shows that many deep learning-based methods have been introduced for the classification of pulmonary tuberculosis using chest radiographs. In this study, we introduced a modified convolutional neural network (CNN) model comprised of 21 layers having different pooling layers, dropouts, and fully connected layers. The objective of this study is to classify images into TB and Normal. The presented methodology is trained and evaluated on the combined dataset formed using four publicly accessible standard datasets. Results of the proposed modified CNN model have been compared with the seven individual pre-trained state-of-art, including VGG16, VGG19, ResNet50, InceptionV3, Xception, MobileNetV2, and EfficientNetB7. This comparison demonstrates that the modified CNN model attained outstanding performance than pre-trained models in terms of accuracy, precision, recall, F1 score, and AUC, which are 0.9081%, 0.9317%, 0.9323%, 0.9307%, and 0.9474%, respectively.