153. APPLICATION OF RAMAN SPECTROSCOPY AND MACHINE LEARNING IN THE CLASSIFICATION OF ESOPHAGEAL SQUAMOUS CARCINOMA

Abstract

Abstract Esophageal squamous cell carcinoma (ESCC) is a common disease among all malignant tumors. Advanced detect technologies of the esophageal cancer, Raman spectroscopy have then been developed with the aim of improving diagnosis sensitivity and specificity and accuracy. The aim of this study was to demonstrate an effective and noninvasive technique, Raman spectroscopy, which was able to differentiate and classify ESCC cell lines (esophageal cancer tissue types). 5 ESCC cell lines and tissues of ESCC patient with staging of T3N1M0 (n = 4) and T3N2M0 (n = 4) in low and high differentiation were examined through Raman spectroscopy. Raman spectra data analysis with four machine learning algorithms: PCA-LDA, PCA-XGB, PCA-SVM and PCA- (LDA, XGB, SVM)-stacked GBM were applied. The Student’s t test was used to analyze the differences between the two groups. PCA-LDA model achieved predictive sensitivity greater than 89%, specificity greater than 96% and the overall accuracy rate greater than 90% for the classification of ESCC cells. PCA-XGM and PCA-SVM models showed overall predictive accuracies of 85% and 87.7%, respectively. The overall accuracy reached to 94.7% in classifying the ESCC cells and normal esophageal cells. In ESCC tissue, Raman intensity of DNA/RNA (780 cm − 1) and lipids (1440 cm − 1) were higher in ESCC tissues compared with that in adjacent tissues. We used PCA-XGB model to classify ESCC tissues and adjacent tissues, which showed the overall predictive accuracy of 85%. This study illustrates Raman spectral traits of ESCC cell lines and esophageal tissues related to clinical pathological diagnosis, and one more possible detection method has been added in the field of esophageal cancer detection. Future studies to determine the role of different Raman spectral features altered in ESCC pathogenesis are warranted.