Raman Spectroscopy and Machine Learning-based Optical Sensor for Rapid Tuberculosis Diagnosis via Sputum

TL;DR

This paper presents a portable Raman spectroscopy-based optical sensor combined with machine learning for rapid, non-invasive tuberculosis detection from sputum samples, achieving high accuracy and potential for point-of-care diagnosis.

Contribution

It introduces a novel portable optical sensor using Raman spectroscopy and PCA-based machine learning for fast TB diagnosis from sputum, outperforming traditional methods.

Findings

100% accuracy for true-positive detection

93.4% accuracy for true-negative detection

Correctly identifies patients on TB medication

Abstract

Tuberculosis (TB) is a contagious disease that causes 1.5 million deaths per year globally. Early diagnosis of TB patients is critical to control its spread. However, standard TB diagnostic tests such as sputum culture take days to weeks to produce results. Here, we demonstrate a quick, portable, easy-to-use, and non-invasive optical sensor based on sputum samples for TB detection. The probe uses Raman spectroscopy to detect TB in a patient's sputum supernatant. We deploy a machine-learning algorithm, principal component analysis (PCA), on the acquired Raman data to enhance the detection sensitivity and specificity. On testing 112 potential TB patients, our results show that the developed probe's accuracy is 100% for true-positive and 93.4% for true-negative. Moreover, the probe correctly identifies patients on TB medication. We anticipate that our work will lead to a viable and rapid TB diagnostic platform.