Rastreio de tuberculose em tempo real por análise salivar através de FTIR e inteligência artificial

Abstract

Introduction: Early diagnosis of tuberculosis constitutes a clinical challenge to be overcome to break the disease transmission chain. New rapid diagnostic methods are needed to eliminate this global public health problem, which mainly affects neglected populations with limited access to healthcare resources. Objectives: Demonstrate proof of concept of the usefulness of salivary analysis through Fourier Transform Infrared Spectroscopy (FTIR) coupled to algorithms analysis of absorption curves by artificial intelligence for the diagnosis of TB in real-time. Material and methods: We analyzed the infrared absorption curves, obtained by FTIR, of 35 patients with pulmonary tuberculosis and 50 healthy, using artificial intelligence algorithms (Random Forest, Gradient Boosting e Neural Network) to distinguish sick and healthy patients. Results: The salivary absorption curves obtained by FTIR, analyzed by artificial intelligence, successfully distinguished healthy individuals from those with tuberculosis with a sensitivity of 80%, specificity of 90% and accuracy of 85%. Conclusion: Salivary analysis by FTIR, coupled with Neural Network, constitutes a possible tool for screening individuals with tuberculosis. This high-performance biophotonic platform has the advantages of needing a small amount of biological sample, low cost, greater biosafety, lower cost, minimum waste production and real-time results.