Discrimination of Dengue Diseases in Children Using Surface-Enhanced Raman Spectroscopy Coupled with Machine Learning Approaches.

This study introduces a novel approach to dengue diagnostics by leveraging surface-enhanced Raman spectroscopy (SERS) coupled to machine learning. This method addresses the critical need for rapid and accurate identification of dengue virus (DENV) infection and prediction of the disease severity. For the first time, a commercialized SERS substrate is applied to analyze plasma samples from 60 pediatric patients, equally distributed among other febrile illnesses (OFI), dengue fever (DF), and dengue hemorrhagic fever (DHF) cases. This innovative application of SERS technology captures unique molecular signature characteristics of each disease state, offering a new paradigm in viral diagnostics. Our methodology employs various machine learning algorithms, notably linear discriminant analysis (LDA) and logistic regression, to classify the SERS spectral data. The models exhibited exceptional performance in distinguishing dengue from OFI, with both achieving an outstanding area under the curve (AUC) of 0.99. In the more complex task of discriminating between DF, DHF, and OFI, LDA demonstrated remarkable AUC values of 0.81, 0.90, and 0.99, respectively, while logistic regression slightly outperformed with AUC values of 0.82, 0.88, and 0.99. Even in the challenging differentiation of DF from DHF, the models achieved notable AUC values of 0.84 (LDA) and 0.79 (logistic regression). This pioneering SERS-based approach represents a significant advancement over existing dengue diagnostic methods, offering unparalleled speed and accuracy, particularly in resource-limited settings. By providing a new tool for early detection and classification of dengue severity, this innovative technique has the potential to improve patient outcomes and guide targeted therapeutic strategies in dengue management.