Prediction of Survival Time of Patients With Esophageal Squamous Cell Carcinoma Based on Univariate Analysis and ASSA-BP Neural Network

• Main Authors: Yanfeng Wang, Enhao Liang, Xueke Zhao, Xin Song, Lidong Wang, Junwei Sun
• Format: Article
• Language: English
• Published: IEEE 2020-01-01
• Series: IEEE Access
• Subjects: Esophageal squamous cell carcinoma; univariate analysis; correlation analysis; back propagation; adaptive salp swarm algorithm
• Online Access: https://ieeexplore.ieee.org/document/9210564/

Esophageal squamous cell carcinoma (ESCC) is one of the most common malignant tumors in the world. In order to find out the influencing factors, univariate Cox regression analysis is used to analyze the blood indexes to screen out the factors affecting the survival or death of patients. Spearman and Pearson correlation analysis can verify whether screening factors are related to survival. The survival curves and progression-free survival curves of 5 factors are given after the threshold is obtained by receiver operating characteristics (ROC). In order to optimize the survival accuracy of patients with ESCC, in view of the low convergence accuracy and easy to fall into local optimization of back propagation (BP) prediction network, the improved adaptive salp swarm algorithm (ASSA), genetic algorithm (GA) and back propagation (BP) neural network are combined. The initial weight and threshold of BP neural network are determined, and the ASSA-BP prediction model and GA-BP prediction model are established. In order to show the reliability and accuracy of the new model in a large range, the ASSA-BP model, GA-BP model and BP model are evaluated respectively. The ASSA-BP model is more effective in predicting the survival time of patients with ESCC, and shortens the training time and improves the prediction accuracy.