A novel improved SMA with quasi reflection operator: Performance analysis, application to the image segmentation problem of Covid-19 chest X-ray images

Slime mold algorithm (SMA) is a meta-heuristic optimization technique based on nature's slime module oscillation modes. Like other meta-heuristic algorithms, SMA is prone to poor diversity, local optima, and imbalanced exploitation abilities. A new, quasi-reflected slime mold (QRSMA) method, which combines the SMA algorithm with a quasi-reflection-based learning mechanism (QRBL), is presented to increase SMA's performance. The enhancement contains two parts: firstly, the QRBL mechanism was established to boost population variety early. Then, quasi-reflection-based jumping (QRBJ) was added to enhance convergence and avoid local optimum in each population update and to maintain the balance between exploitation and exploration. On the CEC20 benchmark functions of various kinds and dimensions, the performance of QRSMA was evaluated and checked that the proposed QRSMA's more robust search capabilities compared to the classic SMA and different search methodologies in terms of statistical, convergence, and diversity measurement. The findings reveal that QRSMA can significantly increase the convergence speed and solution precision of the basic SMA and others by comparing it with basic SMA and other algorithms. Two further tests have also been conducted to assess QRSMA performance. The first is the division of 10 natural gray pictures. Next, the QRSMA was evaluated for a real-world application, such as COVID-19 X-ray images. The region of interest inside the picture containing the characteristics of COVID-19 must be extracted to increase the precision of the classification. Four X-ray images have thus been utilized to assess QRSMA's performance. To evaluate the quality and performance of the QRSMA, comprehensive comparisons have also been carried out using different approaches. Overall test findings show that the QRSMA is an effective Multi-Level Thresholding (MLT) strategy superior to other current methods. (c) 2022 Elsevier B.V. All rights reserved.