Fig. 3: Raman spectroscopic analysis of urine samples for sex-classification.
Median resonance Raman spectra with the respective standard deviation of female (orange, a1, n = 8) and male (blue, a2, n = 9) urine from healthy volunteers using λexc. = 244 nm as the excitation wavelength. Different classification algorithms, i.e., LDA (b) and SVM with linear (c) and radial (d) kernels, are applied, and their respective histograms with smoothed kernel distribution are shown. LDA and linear SVM show similar results with accuracies of approximately 75%. Since this classification problem seems to be nonlinear, the radial kernel SVM can correctly classify both groups. Additionally, the threshold-averaged (solid) ROC curves of all algorithms (e) show the same behavior for the LDA (blue, e1) and SVM [linear (orange, e2) and radial (black, e3)] classifiers. The solid lines represent the mean values, while the shaded areas correspond to the standard deviation, calculated from n = 20-fold cross-validation.
