A novel dynamic Nelder-based Electric Eel Foraging algorithm for global optimization and pathological colorectal cancer image segmentation.

Colorectal cancer (CRC) is a major global health concern, where timely and precise diagnosis is crucial for effective treatment. In medical imaging, accurate segmentation of pathological regions is essential for guiding diagnostic decisions and treatment strategies. However, traditional metaheuristic-based segmentation methods often face challenges like slow convergence, suboptimal threshold determination, and inadequate balancing between exploration and exploitation, which can limit their effectiveness in multi-threshold image segmentation (MTIS) of CRC pathology images. In this study, we propose the Dynamic Adaptive Nelder Electric Eel Foraging Optimization (DANEEFO) algorithm, an enhanced version of the Electric Eel Foraging Optimization (EEFO) algorithm, specifically designed to address these challenges in medical image segmentation. DANEEFO incorporates four strategies to enhance its performance for CRC pathology image segmentation: Latin Hypercube Initialization (LHI), which ensures a structured and diverse population at the start of the search process; the Adaptive Attraction Strategy (AAS), which dynamically balances exploration and exploitation to prevent premature convergence; the Adaptive Nelder-Mead Simplex (ANM) method, improving both global and local search capabilities; and the Dynamic Spread Strategy (DSS), which promotes search diversity to avoid stagnation in local optima. The DANEEFO algorithm is rigorously tested on CEC2017 benchmark functions and is applied to MTIS of CRC pathology images using 2D Renyi entropy and 2D histograms. The experimental results demonstrate that DANEEFO surpasses several classical and recent metaheuristic-based segmentation algorithms in terms of segmentation accuracy and efficiency. Specifically, the proposed method achieves a Peak Signal-to-Noise Ratio (PSNR) of 29.9410, a Feature Similarity Index Measure (FSIM) of 0.9882, and a Structural Similarity Index Measure (SSIM) of 0.9782. These results demonstrate DANEEFO's ability to deliver superior segmentation performance, making it a promising tool for CRC diagnosis and facilitating more accurate treatment planning.