Preliminary Exploration Of Surface-enhanced Raman Spectroscopy For The Prediction Of Early Postoperative Recurrence Of Gastric Cancer

Background:Gastric cancer（GC）is a common malignant tumor of the gastrointestinal tract.It has a high incidence and mortality rate.Although surgery-based comprehensive treatment has improved the prognosis of gastric cancer patients,their postoperative recurrence rate is still high.Therefore,identifying high-risk patients is important for adjusting treatment plans and improving patient survival.The common tools for prognostic assessment of gastric cancer is based on TNM staging,but the prognostic information provided is inadequate.Surface-enhanced Raman scattering（SERS）technique is a method to detect the molecular structure of substances.The combination of SERS and machine learning has great research potential in cancer screening,prognosis prediction and efficacy assessment.Objective:To find the characteristic Raman shift associated with early recurrence of gastric cancer using SERS technique.Search for possible relapse-related markers and try to establish a predictive model for early recurrence of gastric cancer.Method:Post-operative gastric cancer patients were divided into recurrence and non-recurrence groups based on follow-up results.The characteristic peaks were sought by comparing the differences in the preoperative serum SERS spectra of the two groups and inferring their attributed substances based on previous literature.The model was fitted using Principal Component Analysis-Linear Discriminant Analysis（PCA-LDA）and Lasso regression combined with Support Vector Machine（SVM）.Tumour markers or peripheral blood parameters significantly associated with recurrence were included in the SVM model for increasing model prediction capability.The k-fold cross-validation method was used for internal validation of the models.Result:Patients who underwent radical gastric cancer surgery at the Department of Gastrointestinal Surgery,Sichuan Provincial People’s Hospital from April 2021 to July2022 were collected.A total of 152 patients were included,32 in the recurrence group and 120 in the non-recurrence group.Normalizing the spectra revealed that the relative intensities varied significantly at displacements of 500 cm^-1,514 cm^-1,594 cm^-1,726 cm^-1,738cm^-1 and 1543cm^-1.Among them,726 cm^-1 and 1543 cm^-1 Raman peaks were independent risk factors for early recurrence of gastric cancer after surgery,and the substances attributed to them have the potential to be used as recurrence-related markers.Predictive models and k-fold cross-validation using different machine learning modeling approaches.the sensitivity,specificity and accuracy of the model built by PCA-LDA were25%,91.66%and 76.32%,and the AUC of the model ROC was 0.674.the sensitivity,specificity and accuracy of the model built by Lasso combined with SVM were 50%,90%and 81.57%,and the AUC of the model ROC was 0.738.Logistic regression showed that PLR and PNI were significantly associated with postoperative recurrence of gastric cancer.Fitting them into the Lasso-SVM model,the sensitivity,specificity and accuracy of the new model were 56.25%,85%and 78.94%,and the AUC of the model ROC was0.764.Conclusion:The analysis of preoperative serum SERS spectra showed some characteristic peak differences between the recurrence group and the non-recurrence group,which have the potential to be used as recurrence-related markers.The machine learning model based on serum SERS has some potential for predicting early recurrence of gastric cancer after surgery.