Multilevel Chest X-Ray Thresholding Using a Hybrid Metaheuristic Framework

Multilevel thresholding partitions medical images into intensity regions. However, increasing threshold counts challenges the exploration–exploitation balance of metaheuristic algorithms, necessitating efficient solutions for clinical pre-processing. This study develops a sequential hybrid of Sand Cat Swarm Optimization (SCSO) and the Grey Wolf Optimizer (GWO) for normal and COVID-19 chest X-ray thresholding. Operating under a strict 60:40 iteration budget, the best SCSO solution initializes a single GWO individual while randomizing the remaining population, aiming to maximize spatial diversity without computationally heavy operators. Evaluated across two to six thresholds using an Otsu-based objective, SCSO-GWO achieves the most efficient equal-weight mean PSNR (28.3357 dB) and SSIM (0.5963) among seven comparators. Structural efficiency at higher complexities is mathematically validated by Holm–Bonferroni corrected post hoc tests. Furthermore, the mean execution time is reduced to 0.8305 s, representing a 10.66% speedup over the closest hybrid competitor (RSA-SSA) and over 40% against standalone methods. These findings establish SCSO-GWO as a robust, highly time-efficient clinical dimensionality reduction method, offering an optimal quality–time balance to facilitate subsequent medical image analysis workflows.