Detection of Tuberculosis, COVID-19, and Pneumonia from X-ray Images Using CNN

Abstract

Tuberculosis, COVID-19, and pneumonia are lung diseases that require early detection support based on chest X-ray imaging. This study aimed to develop a detection system based on Convolutional Neural Network (CNN) to classify chest X-ray images into normal or disease conditions. This research applied a system development approach using secondary datasets from Kaggle, image processing through size and pixel-intensity normalization, CNN model training using Teachable Machine, and web-based application implementation using Python, Streamlit, OpenCV, Keras, and NumPy. The dataset was divided into 70% training data and 30% testing data. The evaluation results showed model accuracies of 99.37% for Tuberculosis, 100% for COVID-19, and 95.45% for Pneumonia. Application sample testing indicated that all test images were predicted according to the actual diagnosis, with average confidence scores of 98.88% for Tuberculosis, 100% for COVID-19, and 96.66% for Pneumonia. This system may function as a predictive tool for early chest X-ray screening; however, the results must be confirmed by health professionals and further validated using real clinical data.