Among Cardiovascular Diseases (CVD), stroke is the second most prevalent. According to the World Health Organization, 6.7 million individuals suffered strokes in 2017, accounting for 31% of all disease related fatalities globally [1].

17.7 million People died from CVDs in 2017, according to this estimate. Individuals may visit a hospital for a thorough physical examination to determine their risk of having a stroke. Blood pressure, ECG, vascular ultrasonography, and digital subtraction angiography are only a few of the specific assessment elements. To identify stroke risk, plaque image analysis based on image segmentation technology has also been investigated [2]. Hemorrhagic stroke and ischemic stroke are both types of stroke. In China, the primary setting of this study, ischemic stroke occurs 60% to 80% of the time. By taking proactive measures, the diagnosis of ischemic stroke risk aims to lower or avoid the frequency of clinical episodes and early mortality linked to ischemic stroke. The lack of systematic advice for FS when developing stroke risk detection models, which is crucial to the effectiveness of such models, is a major weakness of the research that has been done thus far on assessing the risk of stroke. The selection of prognostic factors in earlier studies was primarily ad hoc, and the most recent findings in medical research were not taken into account. In the past ten years, many computer-aided methods and tools have been created to quickly identify brain disorders. Artificial Intelligence (AI) and Deep Learning (DL) are being used to generate precise and automated findings for stroke diagnosis.

These findings don’t contribute to the detection process on their own. Computer-aided techniques can accompany medical professionals in the process of detecting strokes, which is a very delicate treatment process. The primary research question for this work is therefore: How many significant risk factors should be chosen to be predictive features in a risk detection model for ischemic stroke? In a variety of applications across several other healthcare systems, machine Learning (ML) has rapidly grown and altered. Early identification is crucial for efficient stroke treatment, and ML can be highly beneficial in this process. Throughout the past few decades, there have been a large number of studies on the improvement of stroke diagnosis using ML in terms of accuracy and speed.

ML is an essential area in AI. Algorithms that can learn from the past and get better over time without being explicitly programmed are the subject of ML. There are many different subtypes of ML, but for the sake of this discussion, we will concentrate on supervised learning. Unsupervised learning and DL are also important subtypes. By supervised learning, a model is developed that converts inputs into outputs based on observations and forecasts of the results. It is divided into two categories: regression and classification. Using predictors, classification involves categorizing discrete target variables. Other classification algorithms include k-Nearest Neighbor (kNN) [3], Logistics Regression (LR) [4], Naive Bayes (NB) [5], and Support Vector Machines (SVM) [6]. Regression examines the connection between a target variable’s numerical value and its predictors. Due to their constraints and numerous variables, many real-world issues have broad dimensions and are categorized as NP-hard problems. To deal with such difficult problems, various optimization techniques, including accurate and approximate ones have been devised. Due to the lack of precise optimization techniques, approximation algorithms have been introduced as a novel strategy for managing high-dimensional and multi-state complicated issues [7], [8], and [9]. The two categories of approximate algorithms are heuristic and meta-heuristic algorithms. Heuristic algorithms received less attention due to their limited confinement and application to certain optimization challenges. Recent years have seen a significant increase in the use of meta-heuristic algorithms to solve the majority of challenging real-world multiplexing and nonlinear optimization issues. While meta-heuristic algorithms can deliver workable answers in a reasonable amount of time, they cannot ensure that the best possible solution will be found for a given optimization issue [10]. Investigations over the past two decades have shown several attempts to solve NP-hard problems using meta-heuristic algorithms that are based on the pattern of natural events, animal and human social behavior, and an approximation method. To efficiently explore the solution space for the right solution to an optimization issue, there exist techniques based on a stochastic and approximation approach. There are ways to escape local optimum spots using these procedures. Nonetheless, they are more flexible than heuristic algorithms and can be used for a variety of problems. They are known as black-box optimizers because of this. The dynamic balancing of the two essential components of exploration and exploitation tactics, however, presents considerable difficulty for meta-heuristic algorithms. Exploitation refers to the thorough search in the response space. In contrast, exploration refers to the productive use of the knowledge acquired during the search process and the concentration on the response space’s more promising regions.

References:

• [1] W. M. P. P. Investigators, et al., The World Health Organization Monica Project (monitoring trends and determinants in cardiovascular disease): a major international collaboration, Journal of clinical epidemiology 41 (2) (1988) 105–114.
• [2] U. R. Acharya, M. R. K. Mookiah, S. Vinitha Sree, D. Afonso, J. Sanches, S. Shafique, A. Nico443 laides, L. M. Pedro, J. Fernandes e Fernandes, J. S. Suri, Atherosclerotic plaque tissue characterization in 2d ultrasound longitudinal carotid scans for automated classification: a paradigm for stroke risk assessment, Medical & biological engineering & computing 51 (2013) 513–523.
• [3] L. E. Peterson, K-nearest neighbor, Scholarpedia 4 (2) (2009) 1883.
• [4] R. E. Wright, Logistic regression.
• [5] G. I. Webb, E. Keogh, R. Miikkulainen, Naïve bayes., Encyclopedia of machine learning 15 (2010) 713–714.
• [6] M. A. Hearst, S. T. Dumais, E. Osuna, J. Platt, B. Scholkopf, Support vector machines, IEEE Intelligent Systems and their applications 13 (4) (1998) 18–28.
• [7] F. 452 S. Gharehchopogh, H. Shayanfar, H. Gholizadeh, A comprehensive survey on symbiotic organisms search algorithms, Artificial Intelligence Review 53 (2020) 2265–2312.
• [8] A. B. Gabis, Y. Meraihi, S. Mirjalili, A. Ramdane-Cherif, A comprehensive survey of sine cosine algorithm: variants and applications, Artificial Intelligence Review 54 (7) (2021) 5469–5540.
• [9] H. Zamani, M. H. Nadimi-Shahraki, A. H. Gandomi, Qana: Quantum-based avian navigation 457 optimizer algorithm, Engineering Applications of Artificial Intelligence 104 (2021) 104314.
• [10] B. Abdollahzadeh, F. S. Gharehchopogh, A multi-objective optimization algorithm for feature selection problems, Engineering with Computers 38 (Suppl 3) (2022) 1845–1863.