Predictive modeling of gastric cancer progression: an advanced approach for early detection of plasma membrane invasion

Plasma membrane invasion indicates advanced gastric cancer (GC) with higher metastasis risk, making early prediction crucial. We propose bMSAO-CART, combining memory-guided search artemisia optimization (MSAO) and classification and regression tree (CART) for prediction. MSAO retains high-quality early solutions to accelerate convergence and was validated against 22 algorithms in IEEE CEC 2017 benchmarks. Using data from 1145 GC patients (681 without, 464 with invasion), bMSAO-CART achieved 92.406% accuracy and 94.270% sensitivity. Six important features were identified: serum chloride, total bilirubin, absolute lymphocyte count, alkaline phosphatase, glucose, and triglycerides.