Leveraging Deep Learning Techniques for Tuberculosis Detection from X-Ray Images

This paper presents a thorough approach to developing a tuberculosis detection system with the help of deep learning algorithm on chest X-ray images. In this work, pretrained convolutional neural network models are applied to analyse their performance in detecting tuberculosis. It includes a preprocessing stage of resizing, denoising, normalisation, and data augmentation to strengthen the property of images and their compatibility with the models. The architectures detailed are of ResNet50, VGG-19 deep convolutional layers, and the made-from-scratch custom CNN structure-emphasising each strength in that it is utilised within image classification. The models are evaluated based on ROC curves, precision-recall analysis, and F1 scores to get a balanced prospect of the model's performance. This makes this VGG-19 model of high accuracy and stable reliability. This is further followed by designing a web application integrated with the trained VGG-19 model, where X-ray images can be uploaded for the recognition of TB by end users. The human-friendly interface will also help medical professionals and general users obtain explicit diagnostic results, hence increasing the applicability of deep learning in healthcare diagnostics. This project can transform the detection and treatment outcomes of diseases by making massive contributions to global health using advanced deep-learning techniques in medical diagnosis.