Qualitative Identification of Sialic Acid Content Based on Terahertz Time-Domain Spectroscopy

In this paper, we propose a discrimination model for abnormal sialic acid content using terahertz time and frequency-domain features of samples in an attempt to explore the significance of human serum sialic acid content in early cancer screening. Specifically, the time-domain spectral features were obtained by applying variational mode decomposition (VMD) for time-domain spectrum preprocessing and composite multi-scale entropy (CMSE) for feature extraction. The frequency-domain spectral features were acquired using principal component analysis (PCA). The model was then constructed on the basis of support vector machine (SVM). To verify the reliability of this discrimination model, we used three different concentrations of albumin solutions to simulate the serum environment in the human body and mixed them with sialic acid solutions of eight different concentrations. The results indicate that our discrimination model achieved an accuracy of 91.01% for the detection of abnormal sialic acid content. This study provides a foundation for the accurate quantitative detection of biomarkers in human serum.