Label-Free Classification of a Nasopharyngeal Carcinoma Tissue Test at Different Stages Based on Raman Spectroscopy

Abstract

Abstract Raman spectroscopy (RS) of nasopharyngeal carcinoma (NPC) tissue provides substantial biomolecular information and various biomedicine features for tissue at different stages of cancer development. This study suggested an automatic and quick method for the classification of Raman spectra at different stages of NPC by multivariate statistical analysis. During RS measurement, Raman spectra were acquired from all NPC tissues in two groups of samples: 30 early-stage NPC patients (stages I and II) and 46 advanced-stage NPC patients (stages III and IV). In addition, a tentative diagnostic algorithm comprising principal components analysis and support vector machine was used to effectively classify multivariate data from the Raman spectra to yield sensitivities (70%; 21 of 30 samples) and specificities (91%; 42 of 46 samples) by the leave-one-out cross-validation method. Meaningful chemical compositions in the classification process were then deduced by analyzing the classified mathematical model. This beneficial work provides a great potential clinical method for the automatic classification of NPC stages and the speculation of the chemical compositions for NPC staging.