Rapid and Noninvasive Detection of Brucellosis in Human Based on Serum Fluorescence Spectrum Combined With Machine Learning Algorithms

Abstract

Brucellosis is a notable zoonotic disease caused by Brucella that is often overlooked. Diagnosis involves both clinical symptoms and serological examinations, which are accurate but time-consuming. Therefore, a simple and accurate method is needed. This study aims to assess the potential for diagnosing human brucellosis using serum fluorescence spectra in conjunction with principal component analysis–linear discriminant analysis (PCA-LDA), linear support vector machine (linear SVM), medium radial basis function support vector machine (RBF SVM), K-nearest neighbors (KNN), and decision tree (DT). The study of serum fluorescence spectra in brucellosis-infected compared to healthy revealed that patients with brucellosis had reduced peaks at 452, 624, and 688 nm and elevated peaks at 495 and 643 nm. SVM (linear/RBF) provides more accurate classification results than other algorithms. The method achieved an overall diagnostic accuracy of 89.0%. In conclusion, the serum fluorescence spectrum paired with the SVM (linear/RBF) algorithm is highly promising for human brucellosis detection.