A Modified Gorilla Troops Optimisation-Based Multi-Threshold CT Image Segmentation

This is the authors' abstract. We don't add key points for this paper.

Automated Medical image segmentation plays an important role in clinical diagnosis as it allows pathological area to be precisely located in computed tomography (CT) images. Yet, most of the existing metaheuristic optimization algorithms tend to prematurely converge, having weak local searching ability, and the accuracy of segmentation decreases significantly when the medical image is affected by noise or having poor contrast. A Modified Gorilla Troops Optimisation (MGTO) algorithm is proposed in this paper, which improves the original Gorilla Troops Optimiser (GTO) by means of adaptive exploration–exploitation scheme, enhanced local refinement procedure along with a diversity-preserving search mechanism to obtain a stable convergence and robust performance of the optimizers. The performance of MGTO is initially tested on 23 standard benchmark functions, attaining the best or near-best solutions for 21 functions and yielding an overall average improvement of 44.9% over the original Gorilla Troops Optimiser on multimodal optimization problems. The proposed method is then tested on Kapur entropy-based multi-level thresholding of COVID-19 chest CT image. Experimental results reveal that MGTO consistently achieves better or comparable performance than nine state-of-the-art optimization algorithms in terms of lower Standard Deviation Ratio, higher Peak Signal-to-Noise Ratio, Structural Similarity Index, Feature Similarity Index, and Dice Similarity Coefficient, indicating enhanced segmentation accuracy and structure preservation. Moreover, MGTO has the least processing time, and the computing time is saving 19–40% compared with other methods and the convergence behaviors are stable. The results demonstrate that MGTO can serve as a computationally efficient and accurate optimization paradigm for medical image segmentation, and it is expected to have a wide range of diagnostic imaging applications.