Survival Prediction Model for Patients with Esophageal Squamous Cell Carcinoma Based on the Parameter-Optimized Deep...

Area of Science:

• Oncology
• Machine Learning
• Bioinformatics

Background:

• Esophageal squamous cell carcinoma (ESCC) presents high global incidence and mortality rates.
• Accurate survival prediction models are crucial for improving patient outcomes and clinical management.
• Existing deep belief network (DBN) models for survival prediction are sensitive to parameter selection.

Purpose of the Study:

• To develop a parameter-optimized deep belief network (DBN) model for predicting the survival of patients with esophageal squamous cell carcinoma (ESCC).
• To enhance the performance of the Archimedes optimization algorithm (AOA) by proposing an improved version (IAOA) for DBN parameter optimization.
• To evaluate the efficacy of the proposed IAOA-DBN model against other established prediction models.

Main Methods:

• Feature selection was performed using the Minimum Redundancy Maximum Relevancy (MRMR) algorithm to identify significant survival-associated features.
• A deep belief network (DBN) was employed for survival prediction.
• The Archimedes optimization algorithm (AOA) was adapted to optimize DBN's learning rate (α) and batch size (β), leading to the development of an improved IAOA to address local optima and accuracy issues.
• The performance of the IAOA-DBN model was compared with AOA-DBN, SSA-DBN, PSO-DBN, BES-DBN, IAOA-SVM, and IAOA-BPNN.

Main Results:

• The Minimum Redundancy Maximum Relevancy (MRMR) algorithm successfully identified key features impacting patient survival.
• The proposed Improved Archimedes Optimization Algorithm (IAOA) demonstrated superior performance in optimizing DBN parameters compared to the standard AOA.
• The IAOA-DBN model achieved high accuracy in predicting the five-year survival rate for ESCC patients.

Conclusions:

• The IAOA-DBN model offers a robust and accurate approach for predicting the five-year survival of patients with esophageal squamous cell carcinoma (ESCC).
• This predictive model can serve as a valuable tool to support clinical judgment and treatment planning for ESCC.
• The enhanced IAOA effectively optimizes DBN parameters, improving prediction accuracy in complex medical datasets.