Study on the prognostic model for esophageal cancer survival based on blood indicators and probabilistic membrane system

Yanfeng Wang, Huaiyang Liu, Housheng Li, Suxia Jiang, Junwei Sun

Abstract: Esophageal cancer stands as one of the most common malignant tumors, often with a poor prognosis. The 5-year survival rate for patients with esophageal cancer remains below 30\(\%\). Given the impact of blood factors on esophageal cancer patients, constructing a model to predict their survival duration becomes crucial. This study utilizes the probabilistic membrane system to establish a prognostic survival prediction model for esophageal cancer. Initially, the proc dmzip process step is employed to interpolate missing values, ensuring the integrity and accuracy of the data. Subsequently, for continuous and categorical indicators, ROC analysis and K–M survival analysis are applied, respectively, to assess whether these indicators significantly correlate with the survival of esophageal squamous cell carcinoma (ESCC) patients. Further, COX regression analysis is conducted to determine if the selected indicators are significantly associated with patient survival. Next, a prognostic prediction model tailored to the evolution characteristics of esophageal cancer is constructed using the probabilistic membrane system. The selection of membrane structures and object sets, as well as the design of rules governing the stability, progression, and mortality patterns of esophageal cancer, are carefully considered. Finally, the PMS model is simulated and validated using the MeCoSim software. A comparative analysis of performance metrics is conducted between the PMS model and other prediction models, such as the kernel extreme learning machine (KELM) and the backpropagation neural network (BPNN). The results demonstrate that, compared to KELM and BPNN, the PMS model exhibits higher accuracy and stability.