الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 140 |  |  |
| | | from | 140 | | |
Abstract
Objectives
1)Proposing an improved version of the optimization algorithms inspired by nature, called Heap-based optimizer (HBO), using the Opposition-based learning (OBL) strategy. The efficiency of the proposed algorithm was verified through the use of (CEC’2020). The performance was compared the algorithm is combined with seven other competitive optimization algorithms. The purpose of this optimization is to obtain the best values (Thresholds) using two techniques, Otsu and Kapur, to provide appropriate solutions to face image segmentation problems. The most important of these challenges is the process of dividing images using traditional techniques. These techniques become inefficient, especially when increasing the number of values needed to segment the images.
2)Presenting a new version of optimization algorithms inspired by nature, which is a modern algorithm called Search and Rescue Optimization Algorithm (SAR), and it was integrated with the (OBL) strategy to address problems that may appear in SAR, and the efficiency of the proposed algorithm was verified through the use of (CEC’). 2020), the performance of the algorithm was compared with seven other optimization algorithms inspired by nature. It was then applied to a set of blood images collected at the Barcelona Research Center to detect malaria diseases early in order to address image segmentation problems with two strategies: Otsu and Fuzzy entropy. The results proved that mSAR, with its superior coefficients and main components, exceeds most of the compared algorithms in most cases of where sobriety and retail quality.
3)Evaluating the performance of the mSAR and IHBO algorithms with other optimization algorithms by using three quality measures: signal-to-noise ratio (PSNR), structural similarity (SSIM), and feature similarity (FSIM). The results demonstrated that the proposed improved algorithm (IHBO, mSAR) outperforms competing optimization algorithms. The statistical analysis also shows that the proposed method achieves effective and reliable results compared to other methods.
4)Evaluating the performance of the algorithms proposed in the two theses, IHBO and mSAR, on different types of images in the medical field.
Methodology
1)Despite the developments mentioned in medical image segmentation to detect various diseases in the medical field using naturally inspired optimization algorithms (SIAs), there are many challenges facing the image segmentation process, such as the difficulty of the image segmentation process using traditional techniques. These techniques become inefficient, especially when the number of values needed for segmentation increases. Therefore, an improved version of the IHBO and mSAR algorithm was proposed to provide suitable and fast solutions to medical image segmentation problems.
2)In the original HBO algorithm, search agents are generally subject to the problem of not being able to update locations correctly, which does not allow full use of swarm information or coverage of the entire search space. Also, the initial swarm for HBO is randomly generated, making it impossible to guarantee the initial quality of swarm locations. Therefore, the swarm initialization and location updating mechanisms in HBO are updated as follows.
3) Firstly, we exploit the full information of the entire swarm and using the opposite solutions, OBL is incorporated into the IHBO swarm initialization process. This combination aims to partially expand the search space in the opposite direction of the initial swarm, enhancing the probability of discovering globally optimal solutions and avoiding the trap of local points. Remarkably, all of this is achieved with minimal computational burden.
4) Second, by searching for information within the search space, ES is used as a mechanism to update IHBO swarm positions, with the goal of improving swarm productivity. Local foraging ability and flock diversity are effectively maintained by learning from superior individuals.
5)Finally, a probabilistic survivor selection procedure is implemented for the search agents, giving HBO an exploitative ability by achieving the best solution. At the same time, there is a certain possibility of accepting sub-resolutions in the global resolution areas explored by IHBO.
Results
1)According to distinct environmental conditions in terms of the number of thresholds and the size of the task sets, the performance of (IHBO, mSAR) and the extent of its generalization are verified through experimentation on a set of blood images collected at the Barcelona Research Center to detect malaria diseases early, whether the image is color or black and white.
2)The performance of the improved algorithms (IHBO, mSAR) was compared with a group of other optimization algorithms, namely Slap swarm algorithm (SSA), Moth flame optimization (MFO), Gray wolf optimization (GWO), Sine cosine algorithm (SCA), Harmony search optimization. (HS), Electromagnetism optimization (EO), and Ant lion optimizer (ALO), L’evy flight distribution (LFD), Harris hawks optimization (HHO), equilibrium optimizer (EO), gravitational search algorithm (GSA), arithmetic optimization algorithm ( AOA), and the original SAR using three quality metrics: signal-to-noise ratio (PSNR), structural similarity (SSIM), and feature similarity (FSIM).
3) A set of other metrics were used to evaluate the quality of image segmentation and its efficiency in finding quick and appropriate solutions to image segmentation problems. Among the types of these metrics is the CEC’2020 benchmark functions. Wilcoxon rank test, standard deviation (STD)
4) The results demonstrated that the proposed improved algorithm (IHBO) outperforms competing optimization algorithms. The statistical analysis also shows that the proposed method (IHBO) achieves effective and reliable results compared to other methods in gray level images. As demonstrated by the results of the mSAR algorithm with its hyper parameters and principal components, it exceeds most of the compared algorithms in most cases in terms of robustness and segmentation quality of RGB images.
Recommendations
1)Although this work addresses many of the problems facing medical image segmentation and greatly enhances its overall performance, in future work the IHBO algorithm or the mSAR algorithm could be used and combined with machine learning or deep learning algorithms by incorporating input feature selection and parameter optimization for the methods. Machine learning.
2)Study the performance of hybridizing the original HBO or SAR algorithms with another metaheuristic algorithm to automatically search for the optimal number of thresholds for a given image.
3)There are many SIAs that have not yet been applied for use in segmentation task optimization, so it is necessary to develop other SIAs in this field and measure their efficiency. Moreover, applying SIAs to complex image analysis problems is mandatory.
4)Evaluating the performance of the algorithms proposed in the two theses, IHBO and mSAR, on different types of images.
5)Integrating methods such as (Otsu, Fuzzy entropy, Kapur) that must be developed to test the efficiency of their results on several types of images in different fields.