Diagnosis of Dental Fluorosis Using Micro-Raman Spectroscopy Applying a Principal Component-Linear Discriminant Analysis

Abstract

Dental fluorosis is an irreversible condition caused by excessive fluoride consumption during tooth formation and is considered a public health problem in several world regions. The objective of this study was to evaluate the capability of micro-Raman spectroscopy to classify teeth of different fluorosis severities, applying principal component analysis and linear discriminant analysis (PCA-LDA), and estimate the model cross-validation accuracy. Forty teeth of different fluorosis severities and a control group were analyzed. Ten spectra were captured from each tooth and a total of 400 micro-Raman spectra were acquired in the wavenumber range of 250 to 1200 cm^-1, including the bands corresponding to stretching and bending internal vibrational modes ν₁, ν₂, ν₃, and ν₄ (PO₄^3-). From the analysis of the micro-Raman spectra an increase in B-type carbonate ion substitution into the phosphate site of the hydroxyapatite as fluorosis severity increases was identified. The PCA-LDA model showed a sensitivity and specificity higher than 94% and 93% for the different fluorosis severity groups, respectively. The cross-validation accuracy was higher than 90%. Micro-Raman spectroscopy combined with PCA-LDA provides an adequate tool for the diagnosis of fluorosis severity. This is a non-invasive and non-destructive technique with promising applications in clinical and epidemiological fields.

Keywords: Raman spectroscopy; dental fluorosis; discriminant analysis; principal component analysis.

Figure 1

Photograph depicting a rectangular area used for Raman spectra acquisition under a digital microscope.

Figure 2

Mean normalized Raman spectra of sound, mild, moderate, and severe dental fluorosis categories.

Figure 3

Histogram of the height distribution of the 960 cm⁻¹ Raman spectrum band by fluorosis category (100 spectra in each category).

Figure 4

Principal component analysis (PCA). Scatter plot in the 3D PCA space of the Raman spectra according to the fluorosis category.

Figure 5

Principal component analysis (PCA). Loadings plot of principal component 1 (PC1) and principal component 2 (PC2) of the Raman spectra data.

Figure 6

Partition plot resulting from the linear discriminant analysis in the 3D principal component space. The colors set the limits of each classification area defined by the LDA model (blue = sound, green = mild, yellow = moderate, and red = severe). The black marks indicate the sample was correctly classified, while red marks indicate the sample was incorrectly classified by the model. The black dots represent the group mean value.