Enhancing Tuberculosis Detection: Leveraging RF-HOG Model for Automated Diagnosis from Chest X-ray Images

Abstract

Tuberculosis (TB) is undeniably an infectious condition that stems from the Mycobacterium tuberculosis bacterium. While it primarily impacts the lungs, it can also harm other body regions if not addressed promptly. TB control in India faces several challenges, including high population density, inadequate healthcare infrastructure in certain regions, limited awareness, stigma associated with the disease, and the emergence of drug-resistant strains. These challenges require ongoing efforts to improve diagnosis, treatment, and prevention strategies. To address the burden of TB, recent advancements in machine learning technologies employed to efficiently detect this disease at an early stage. A completely automated diagnostic system can accurately predict the presence of TB, effectively preventing prolonged lung infections and related diseases caused by TB. In this research, we employed augmentation techniques with HOG (Histogram Oriented Gradients) to extract features from chest X-ray images of TB. These features were then utilized in a Random Forest Classifier to train the proposed model of RF-HOG for classifying Tuberculosis and Normal cases from the image. This proposed method holds promise for diagnosing TB in chest X-ray images and can be beneficial for medical analysts.