Application Of Neural Network Model In Survival Analysis Of Patients With Endometrial Cancer

Endometrial carcinoma(EC)refers to a group of epithelial malignancies that occur on the endometrium.As a common malignant tumor of female genitalia with high morbidity and mortality,it has become a major public health burden.Clinical studies have shown that scientific assessment of the survival status of cancer patients can help doctors make scientific clinical treatment decisions and prognostic judgments,so as to provide patients with more individualized and accurate treatment plans,and ultimately improve the survival rate of patients.However,until now,there has been a lack of an effective method to scientifically evaluate the survival status of patients with endometrial cancer.Most of the previous studies on the prognosis of endometrial cancer focused on risk assessment,while relatively few studies on predicting the survival time of patients.In addition,Cox proportional risk model,decision tree,support vector machine and other binary classification models are more commonly used to evaluate the death risk of patients with endometrial cancer,among which Cox proportional risk model is limited by strict linear assumption.However,binary classification models such as decision tree and support vector machine can not realize continuous prediction of survival conditions of patients at different time points.To some extent,this makes it impossible for us to scientifically evaluate the survival status of patients with endometrial cancer.However,with the continuous development of deep learning technology,this problem has been effectively solved.On the one hand,people have built various neural network models based on Cox proportional risk model,which can not only identify the nonlinear relationship between variables,but also predict the probability of events occurring at different time points.On the other hand,the method of survival time prediction has made a new breakthrough.These model methods have shown excellent performance in many tasks in the past,and provide a methodological basis for us to scientifically evaluate the survival status of patients with endometrial cancer.In this study,a neural network model was established based on the data of endometrial cancer in the SEER(Surveillance,Epidemiology and End Results)database,with the aim of scientifically assessing the survival status of patients.Firstly,we filter the data,perform a simple statistical analysis and divide the training set and the test set.Next,DBP model(Deep Bayesian Perturbation Cox Network)was used to predict the death risk of patients with endometrial cancer.On this basis,the high and low risk groups of patients were divided and the survival rate and survival state were predicted.At the same time,Deep CENT model was used to predict the survival time of patients with endometrial cancer,and the predicted results were compared with the traditional model.The results showed that the prediction accuracy of the neural network model was superior to the traditional model no matter from the perspective of death risk or survival time.In addition,we analyzed the factors affecting the survival of patients with endometrial cancer combined with analysis of variance,and the results showed that the survival status of patients with endometrial cancer was deteriorating with the increase of the age of diagnosis and the development of the disease.Meanwhile,it was found that the two variables of surgery at other or distant metastatic sites and whether chemotherapy had a significant impact on the survival status of patients.Finally,we use the data of a hospital to verify the external space of the neural network model,which proves that the DBP model has strong transferability.