Raman spectroscopy assisted support vector machine: a steadfast tool for noninvasive classification of urinary glucose of diabetes mellitus

The currentstudy presents the application of a Raman spectroscopy-assisted Support Vector Machine (SVM) algorithm for the analysis and classification of urine samples of diabetes mellitus and healthy control subjects. Raman spectra of pathological as well as control samples were acquired in the spectral range of 600 to 1800 cm-1. Spectral differences were observed in Raman spectra of diabetic and non-diabetic urine samples based on variations in their chemical composition. Effective machine-learning techniques were utilized to highlight the spectral differences between diabetic and normal urine samples. Data classification was carried out by using SVM models with different kernel functions: radial basis function (RBF), polynomial function, linear function, and sigmoid. The algorithm classifies data sets into different classes based on even subtle variations in the spectral features. The classification performance of the model was evaluated by employing a 10-fold cross-validation method. The best performance was obtained with RBF having a diagnostic accuracy, precision, sensitivity, specificity, F-score, and AUC of approximately 94%, 100%, 90%, 100%, 0.94, and 0.99 respectively.