| BackgroundPeritoneal metastasis of gastric cancer has a subtle onset and is difficult to detect in the early stages,which is one of the main reasons for poor prognosis.Accurate staging is of great clinical significance for formulating treatment plans.In this study,we aim to develop a multimodal prediction model for peritoneal metastasis of gastric cancer,which combines clinical indicators,radiomics,and the specific expression protein CTSL.The model will be accurate,stable,and interpretable,providing support and guidance for the development of personalized treatments in clinical practice.Chapter Ⅰ:Construction of a Prediction Model for Peritoneal Metastasis in Gastric Cancer by Combining Clinical Indicators and Serum Tumor MarkersMethods:A total of 3,480 gastric cancer patients from Nanfang Hospital affiliated to Southern Medical University and Wuhan Central Hospital affiliated to Huazhong University of Science and Technology were included and divided into an internal training group,an internal validation group,and an external validation group(1,949,704,and 827 patients,respectively).Eleven clinical indicators were studied,including gender,smoking status,tumor size,differentiation degree,Borrmann type,tumor location,T stage,and serum tumor markers(STMs,including CA19-9,CA724,and CEA).Analysis of variance(ANOVA),recursive feature elimination(RFE),and regression analysis were used to select clinical features and establish a clinical prediction model.After statistical analysis,tumor size,Borrmann type,tumor location,T stage,and STMs were selected as the seven indicators to establish a joint prediction model.The area under the ROC curve(AUC)was used to evaluate the model’s predictive ability for peritoneal metastasis,and the decision curve analysis(DCA)was used to evaluate the clinical value of the model.Results:The AUC of the simple clinical indicator model in the internal training group,internal validation group,and external validation group were 0.762,0.772,and 0.758,respectively.After incorporating STMs,the AUC of the model improved to 0.806,0.839,and 0.801 in the same groups.DCA curve analysis demonstrated that the combined model has better clinical value in predicting peritoneal metastasis of gastric cancer.Conclusion:The simple clinical indicator model had a certain predictive ability for peritoneal metastasis in gastric cancer,and the diagnostic performance of the prediction model was significantly improved after combining serum tumor markers,which can better predict peritoneal metastasis.The model is simple and easy to obtain and is worthy of clinical promotion and use.Chapter Ⅱ:Construction of a Gastric Cancer Peritoneal Metastasis Imaging Omics Prediction Model Using the Transformer Method.Methods:Retrospective analysis of CT images of gastric cancer patients from Nanfang Hospital affiliated to Southern Medical University and Wuhan Central Hospital affiliated to Huazhong University of Science and Technology was performed.A Transformer-based model was constructed to predict peritoneal metastasis in gastric cancer.The receiver operating characteristic(ROC)curve was used to evaluate the performance of the model,and the clinical effectiveness of the model was assessed using decision curve analysis(DCA).Results:A total of 48 patients with peritoneal metastasis due to gastric cancer were included in the study,along with 50 non-peritoneal metastasis patients who were randomly matched,making a total of 98 patients in the training group.In addition,153 patients who met the criteria were included in the internal validation group at the affiliated Nanfang Hospital of Southern Medical University.42 patients who met the inclusion criteria were also included in the external validation group at the Gastrointestinal Surgery Department of Wuhan Central Hospital affiliated to Huazhong University of Science and Technology.The gastric cancer peritoneal metastasis prediction model,constructed using Transformer,had an AUC curve area of 0.893(95%confidence interval 0.875～0.934)in the internal validation group,and 0.862(95%confidence interval 0.838～0.924)in the external validation group.Decision curve analysis(DCA)was used to evaluate the clinical efficacy of the gastric cancer peritoneal metastasis prediction model constructed using Transformer,showing good clinical performance in predicting peritoneal metastasis of gastric cancer.Conclusion:The gastric cancer peritoneal metastasis prediction model based on the Transformer method can effectively predict peritoneal metastasis and has significant clinical application value.Chapter Ⅲ.Study on the biological function and molecular mechanism of CTSL in predicting peritoneal metastasis based on bioinformatics analysisMethods:A gastric cancer peritoneal metastasis dataset was searched,and differentially expressed genes(DEGs)related to epithelial-mesenchymal transition(EMT)were identified through bioinformatics analysis.The expression of the identified DEG was validated by immunohistochemistry(IHC),and its association with gastric cancer peritoneal metastasis was analyzed.The effect of regulating the expression of this gene in gastric cancer cells on their invasive ability was assessed,and its role in modulating tumor-associated macrophages(TAMs)in the tumor microenvironment was investigated.Results:Bioinformatics analysis identified six DEGs,and CTSL,a gene related to EMT,was selected for further analysis.CTSL was found to be closely associated with tumor staging and mainly expressed in macrophages.IHC showed that CTSL expression was significantly higher in gastric cancer tissues than in normal tissues(P<0.01).In gastric cancer tissues,CTSL expression was higher at the tumor margin than in the tumor center,and the difference was statistically significant(P<0.01).Immunofluorescence staining showed that CTSL was mainly expressed in M2-type macrophages and induced TAMs to differentiate into M2-type macrophages,thereby promoting gastric cancer metastasis by altering the local tumor microenvironment.Conclusion:CTSL is a DEG in gastric cancer tissues that is closely related to peritoneal metastasis.It can promote peritoneal metastasis by regulating the invasive ability of gastric cancer cells and promoting the differentiation of TAMs into M2-type macrophages.Chapter Ⅳ.Preliminary Study on Multi-Modal Construction of Peritoneal Metastasis Prediction Model in Gastric Cancer.Method:A retrospective analysis was conducted on pathological slides of gastric cancer patients.A gastric cancer peritoneal metastasis prediction model was constructed by combining clinical features(tumor size,Borrmann classification,tumor location,T stage,and seven clinical indicators of STMs),imaging features,and CTSL immunohistochemical expression.Results:The ROCs of the gastric cancer peritoneal metastasis prediction model constructed by combining clinical features and imaging features,as well as CTSL,were 0.931(95%confidence interval,0.898-0.968)and 0.913(95%confidence interval,0.886-0.959),respectively.The decision curve analysis(DCA)showed that the predictive model had good clinical efficacy when combining clinical features,imaging features,and CTSL immunohistochemical expression.Conclusion:The multi-modal gastric cancer peritoneal metastasis prediction model constructed by combining clinical features,imaging features,and CTSL immunohistochemical expression has good predictive ability. |
PDF Full Text Request