An enhanced animated oat optimization framework for accurate MRI image segmentation

Accurate MRI image segmentation using multilevel thresholding is difficult because the threshold search space becomes increasingly multimodal as the number of thresholds grows, while stochastic optimizers may suffer from weak initial diversity and premature stagnation. This paper proposes a targeted enhancement of the Animated Oat Optimization (AOO) algorithm by integrating Opposition-Based Learning (OBL) during initialization and a Local Escaping Operator (LEO) after the AOO position update. OBL improves the initial coverage of the search space, whereas LEO provides controlled perturbation and local refinement to help the population leave deceptive regions. The resulting AOO + OBL + LEO framework is evaluated first on the CEC2022 numerical benchmark suite using 30 independent runs and then applied to unsupervised multilevel Otsu segmentation of diagnostic MRI images. On the CEC2022 experiments, AOO + OBL + LEO obtains the best overall Friedman mean rank of 1.833 among the compared methods, with the best mean fitness on 8 of the 12 benchmark functions. In the MRI case study, the proposed framework is assessed using fitness value, PSNR, SSIM, FSIM, Wilcoxon tests, and runtime trends across multiple datasets and threshold levels. The results show that the enhanced AOO variants provide stable and competitive segmentation quality under the adopted experimental protocol, while the ablation results confirm the complementary effects of OBL and LEO.