A novel detection technology for early gastric cancer based on Raman spectroscopy

Despite universal endoscopic screening, early detection of gastric cancer is challenging, led researchers to seek for a novel approach in detecting. Raman spectroscopy measurements as a fingerprint of biochemical structure, enable accurate prediction of gastric lesions non-destructively. This study aimed to evaluate the diagnostic power of Raman spectroscopy in early gastric cancer (EGC), and reveal dynamic biomolecular changes in vitro from normal to EGC. To clarify the biochemical alterations in Correa's cascade, Raman spectra of human normal gastric mucosa, intestinal metaplasia, dysplasia, and adenocarcinoma were compared at tissue and cellular levels based on a self-developed data processing program. For effectively identify EGC, Raman spectroscopy was used combined with multiple machine learning methods, including partial least-squares discriminant analysis (PLS-DA), support vector machine (SVM), and convolutional neural network (CNN) with leave-one-out (LOO) cross validation. A total of 450 Raman spectra were investigated in this study. The upregulation of ν_sym(O-P-O) backbone (p < 0.001) was identified as a favorable factor for the diagnosis of EGC, the area under the ROC curve (AUC) was up to 0.918. In addition, higher levels of lactic acid (p < 0.001), lipids (p < 0.001), phenylalanine (p = 0.002), and carotenoids (p < 0.001) were detected in EGC. Multivariate machine learning methods for diagnosis of EGC based on Raman spectroscopy, the sensitivity, specificity, accuracy, and AUC were 91.0%, 100%, 94.8%, and 95.8% for SVM, and 84.8%, 92.0%, 88.8%, and 95.5% for CNN, respectively. Raman spectroscopy can be used as a powerful tool for detecting EGC while elucidating biomolecular dynamics in tumorigenesis. (Chictr.org.cn, ChiCTR2200060720.).