Optimized Automated Baseline Correction for NS1 Adulterated Salivary Raman Spectra

Non-structural protein 1 (NS1) is a biomarker for flavivirus infection related diseases. NS1 protein feature can be captured using Surface Enhanced Raman spectroscopy (SERS) technique. SERS spectra of NS1 produced a significant NS1 characteristic peak at 1000 cm⁻¹ Raman shift. However, as the concentration of NS1 reduced, the intensity of the peak is reduced producing low signal to noise ratio (SNR) spectra. In addition, the spectra is also interfered with the baseline drift originated from fluorescence effect produced by the protein itself. Baseline drift is a common unwanted feature of a Raman spectrum and removing it will improve the spectra quality for visual inspection or automated classification. In this study, baseline drift of Raman spectra exhibiting low intensity of NS1 peak is removed using Piecewise Cubic Hermite (PCH) interpolation technique in Matlab environment. Systematic tuning of window size and step size is performed to obtain the optimum baseline removal parameter removing the baseline drift while preserving the characteristic of NS1. By observing the value of SNR and NS1 characteristic peak intensity, I _max, the optimum window size and step size identified to be 40 and 9, respectively.