| Gastric cancer is one of the most common malignant tumors in the world,seriously threatening human life and health.And with poor living habits and the toxic side effects of cancer treatment,the risk of cancer patients dying from cardiovascular diseases has increased dramatically.With the popularity of electronic medical records,the emergence of radiomics,and the rise of personalized medicine,researchers are increasingly expecting to be able to mine influencing factors related to the prognosis of gastric cancer patients from their clinical data and radiomics data,thereby achieving prediction of the survival conditions of gastric cancer patients and providing decision-making for clinical personalized treatment.Therefore,this article has carried out research on prediction models for the prognosis of gastric cancer in order to improve the accuracy of prediction of survival conditions of gastric cancer patients and provide objective basis for clinical treatment decisions.The main research contents of this article are as follows:(1)Research on survival prediction models for gastric cancer patients with cardiovascular disease.First,the random survival forest and stepwise backward regression method were used to screen for risk factors related to cardiovascular disease death in gastric cancer patients.Then,a Cox proportional hazards model was constructed,and the model results were visualized and evaluated using nomograms and forest plot.The 5-year and 10-year survival rates of patients with gastric cancer and the impact of related influencing factors on the prognosis of gastric cancer patients were analyzed.Then,through propensity score matching,chemotherapy and non chemotherapy patients were compared between groups.The results showed that chemotherapy as a protective factor could reduce the risk of cardiovascular death in gastric cancer patients.The proposed method has a C-index value of 0.8044,an IBS value of 0.093,and an AUC value of 0.834 in predicting cardiovascular disease survival in patients with gastric cancer,which is superior to other comparative methods.(2)Research on predictive models for survival status and survival rate of gastric cancer patients based on multimodal data.On the one hand,a GBDT(Gradient Boosting Decision Tree)classification model is constructed by combining the two modality data of computed tomography(CT)imaging radiomics and clinical features to predict the 2-year survival status of gastric cancer patients.The accuracy of the GBDT classification model reached 98%,which is superior to other comparison models;and compared with the classification accuracy of single modal data,the classification effect based on multimodal data is better.On the other hand,a Cox proportional hazards model is constructed based on the above two modal data to predict the survival rate of gastric cancer patients.The results show that the survival rate prediction based on multimodal data is better than that based on single modal data for gastric cancer patients.The above two methods proposed in this article have achieved good performance,can provide reliable treatment basis for doctors,and have certain clinical value. |
PDF Full Text Request