A Study of Spectral Preprocessing Techniques for Bilirubin Quantification Using DRS

Spectroscopic studies have gained significant impact in the medical field due to its non-invasive nature. Bilirubin, a key biological chromophore responsible for jaundice, exhibits a unique absorbing peak around 450 nm in the visible spectrum. Diffuse reflectance spectroscopy has proven effective in quantifying the bilirubin values by analyzing the spectral features. However, the quality of spectral data can be affected by noise, scattering and other environmental conditions, potentially affecting the accuracy of bilirubin estimation. In this work, we applied a few preprocessing techniques to the spectral data and assessed their impact on the accuracy of diffuse reflectance based bilirubin quantification. The results showed that the Multiplicative Scatter Correction method performs well with a mean error of $-0.02 \mathrm{mg} / \mathrm{dL}$, root mean square error of $0.52 \mathrm{mg} / \mathrm{dL}$ and Pearson’s r value of 0.9.