An enhanced spider wasp optimization algorithm for multilevel thresholding-based medical image segmentation
Mohamed Abdel-Basset,Reda Mohamed,Ibrahim M. Hezam,Karam Sallam,Ibrahim A. Hameed
DOI: https://doi.org/10.1007/s12530-024-09614-4
IF: 2.347
2024-09-03
Evolving Systems
Abstract:Early in 2019, COVID-19 was discovered for the first time in Wuhan, China, resulting in the deaths of a significant number of people in many different countries all over the world. Due to the rapid spread of this epidemic, scientists have strived to find quick and accurate diagnostic methods to lessen its global impact. Chest X-ray images were the best tool for rapidly and safely detecting COVID-19, but the manual examination of those images might result in faulty diagnoses. Therefore, the scientists have used deep learning (DL) models to remedy this shortcoming and classify the images infected with COVID-19 more accurately. Image segmentation is an essential step in improving the classification accuracy of DL models. Among existing image segmentation techniques, multilevel thresholding-based image segmentation techniques have gained significant interest due to their simplicity and high accuracy. However, the computational cost of those techniques exponentially increases as the number of threshold levels increases. Therefore, over the last few years, metaheuristic algorithms have collaborated with those techniques to significantly lessen the computational cost and accurately solve the image segmentation problem. However, those algorithms have some shortcomings, such as falling into local minima and slow convergence speed, which make them unable to find precise results. Therefore, in this paper, we present a new multilevel thresholding-based medical image segmentation technique based on the recently proposed spider wasp optimizer (SWO) to better segment the medical images, especially the chest X-ray images for detecting COVID-19 infection more accurately and rapidly. In addition, SWO is enhanced by two newly proposed mechanisms, namely global search improvement and local search improvement, to present a new better variant, namely improved SWO (ISWO). The former mechanism is responsible for improving the exploration operator by sharing the knowledge of the current individual and a newly generated individual, while the latter aims to improve the exploitation operator to improve the convergence speed. To evaluate the stability of ISWO and SWO, ten COVID-19 X-ray images with heterogeneous histograms under nine threshold levels (T) are used. Also, they are compared to eight rival optimizers according to several performance metrics to demonstrate their efficacy. According to the experimental results, ISWO is the best-performing algorithm, followed by SWO. Quantitatively, ISWO could achieve an average fitness value of 2796.837, while SWO could reach a value of 2796.33.
computer science, artificial intelligence