Advanced Tuberculosis Detection System for Preserving Public Health

Abstract

Tuberculosis is a major public health threat that requires efficient and early detection technologies. Challenges associated with real-world implementation include information variability, model availability, and operations in different clinical circumstances. In this paper, we introduce a Crow Search Driven Tuned Logistic Regression (CS-TLR) tuberculosis identification system to sustain early identification of the disease and increase the levels of diagnostic accuracy to help sustain public health. The model attempts to enhance the overall prediction performance of logistic regression by optimizing its parameters by using CSA's optimization capability. We gathered the Kaggle dataset with a variety of clinical and demographic characteristics of tuberculosis patients. We evaluate the performance of the suggested method employing standard parameters including AUC (90.5%), precision (89.9%), recall (90.1%), and F1-score (89.7%). The findings suggest that it could be an important public health surveillance tool by enabling early diagnosis and treatment of TB.