Multi-threshold image segmentation method based on multi-strategy enhanced hunger games search algorithm: case study of breast cancer pathological images

• Published in: Displays Vol. 90; p. 103095
• Main Authors: Long, Zeye, Chen, Ying, Zhang, Yuliang, Lei, Feiyang, Xie, Yudi, Guo, Shubin, Deng, Junkang, Chen, Huiling, Liu, Lan
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.12.2025
• Subjects: Breast cancer; Hunger games search; Multi-threshold image segmentation; Swarm intelligence; Multi-threshold image segmentation; Swarm intelligence; Breast cancer; Hunger games search
• ISSN: 0141-9382
• DOI: 10.1016/j.displa.2025.103095

•A multi-strategy enhanced hunger games search algorithm, named MEHGS, is proposed based on HGS.•Multi-dimensional performance comparisons are conducted between MEHGS and eight traditional algorithms and nine improved algorithms using 30 benchmark functions from the CEC2017 test suite. The results indicate that MEHGS exhibits faster convergence speed and stronger ability to escape local optima.•A novel multi-threshold image segmentation method is proposed by combining MEHGS with non-local mean 2D histogram and Kapur’s entropy.•The proposed multi-threshold image segmentation method is applied to segment breast cancer pathological images, yielding high-quality segmentation results to assist doctors in diagnosis. Breast cancer, as a prevalent form of malignant tumors, presents a significant health risk to women. Therefore, accurate diagnostic support is crucially important. Effective support for diagnosis can be provided by breast cancer pathological image segmentation, and the multi-threshold image segmentation method is widely used due to its simplicity and effectiveness. However, with the increase in the number of thresholds, the computational complexity is gradually increased, which may lead to a decrease in the accuracy of segmentation. To address this issue, a multi-strategy enhanced hunger games search (MEHGS) algorithm is proposed based on the hunger games search (HGS) to determine the optimal threshold. In MEHGS, a competitive mechanism is used to divide the population into excellent subpopulation and common subpopulation, and corresponding individual updating strategies are designed based on the characteristics of subpopulations to enhance the search capability of the algorithm; an adaptive cross-disturbance strategy is introduced to avoid premature convergence. To evaluate its performance, firstly, the CEC2017 test suite is used to conduct multi-dimensional performance comparisons between MEHGS and other advanced algorithms. The results show that, compared to other advanced algorithms, MEHGS demonstrates stronger convergence performance and global search capability. Additionally, a novel multi-threshold image segmentation method is proposed for breast cancer pathological image segmentation by combining MEHGS with non-local mean 2D histogram and Kapur’s entropy. Experimental results indicate that this method outperforms other similar methods in terms of image segmentation performance at both low and high threshold levels, showing great potential in breast cancer auxiliary diagnosis.