Interstitial lung diseases (ILDs) are a heterogeneous group of pulmonary disorders where delayed risk stratification significantly increases mortality. However, accurate early prognosis is often hindered by highdimensional, redundant clinical features and the complex non-linear interactions among them. We propose a hybrid feature selection model for ILD mortality prediction. WBCPSMA, a novel optimizer integrating Comprehensive Learning Particle Swarm Optimization (CLPSO) and the Slime Mould Algorithm (SMA) with adaptive weighting exploration and optimal backtracking guidance, was combined with a Kernel Extreme Learning Machine (KELM) classifier to enable accurate prediction. The method was validated on the CEC2014 and CEC2017 benchmark suites to confirm its global optimization capability and evaluated on a clinical ILD dataset comprising 211 patients and 22 features. To ensure rigorous validation on a limited dataset, we employed a Nested Cross-Validation protocol with Bootstrap resampling. The model achieved an Accuracy of 0.806 (95% CI: 0.742-0.855) and an AUC of 0.836, outperforming seven state-of-the-art methods. Crucially, it demonstrated superior clinical utility by maintaining a balanced Sensitivity (0.798) and Specificity (0.814) via dynamic thresholding, alongside reliable probability calibration (Brier Score: 0.162). The algorithm identified four key biomarkers-HRCT score, PaO2/FiO2, Lactate, and Age-with SHAP analysis confirming their clinical relevance. Furthermore, external validation on four UCI datasets verified the method's robust generalization capability. This approach offers high predictive performance and transparent decision support for ILD mortality risk stratification, providing a foundation for future integration into clinical workflows.

Path planning is vital for Unmanned Aerial Vehicle (UAV), especially in low-altitude environments. This paper proposes Interactive Mechanism and Enhanced Solution Quality-driven Artemisinin Optimization (IEAO) to address UAV path planning challenges. The Interactive Mechanism enhances population diversity through communication among individuals within the population. Furthermore, IEAO combines the Artemisinin Optimization update rule with an Enhanced Solution Quality strategy to refine solution selection, thereby enhancing convergence accuracy. The effectiveness of IEAO was evaluated using IEEE CEC2017 and IEEE CEC2019 test sets. The results show that IEAO outperforms other classical and advanced algorithms in 87.26

The continuous version of the Ant Colony Optimization algorithm (ACOR) is a meta-heuristic method inspired by information exchange in ant colony. While ACOR is widely recognized for its simplicity, ease of implementation, and competitive performance in low-dimensional optimization problems, it faces significant challenges in high-dimensional and complex scenarios. These challenges include low convergence precision, slow convergence speed, and a tendency to fall into local optima. This study proposes a novel robust and adaptive ant colony optimization algorithm called Elite Guidance Ant Colony Optimization (EGACO) to address these limitations. EGACO introduces an elite guidance strategy that allows elite populations to guide the updates of ordinary populations while maintaining diversity and robustness. Additionally, EGACO replaces Gaussian sampling with adaptive t-distribution sampling, dynamically adjusting the degrees of freedom to ensure a more balanced search range. To comprehensively evaluate the performance of EGACO, we conducted extensive experiments on CEC2017 benchmark functions with 30, 50, and 100 dimensions, comparing it against ACOR and 15 other state-of-the-art algorithms. Furthermore, EGACO was applied to multi-threshold segmentation of colon cancer histopathology images. The results show that EGACO has a significant advantage over ACOR on 29 of the CEC2017 functions (p ≤ 0.05) and maintains superior performance compared to other leading algorithms. It also leads in both quantitative and qualitative segmentation results, highlighting its potential for solving complex real-world problems.

Tubular structures, such as pulmonary and cerebral blood vessels, serve as crucial biomarkers for understanding disease mechanisms and for diagnostic and therapeutic purposes. However, current vessel segmentation methods often struggle with suboptimal accuracy. For example, in three-dimensional medical imaging, the imbalance between boundary and non-boundary voxels in vascular structures makes it difficult to capture the relationships between adjacent slices, leading to discontinuities in the vascular segmentation predictions. As such, learning correlations across different spatial depths and extracting global features have become central research goals. To address these challenges, we propose a novel deep learning model, DWCS-Net, designed to enhance vascular segmentation in medical images. DWCS-Net integrates Depth-Wise Fusion Attention (DWA) and ConvNeXt blocks, enabling the effective capture and integration of multi-scale global features while learning the correlations between different spatial dimensions. This capability strengthens the model's ability to detect complex geometric structures and addresses the limitations of existing approaches. Experimental results on the PARSE22 dataset demonstrate that DWCS-Net improves the DSC by 0.82% over nn-UNet, reduces HD95 by 0.15 mm, and outperforms MedNeXt by 0.71% in DSC. On the CAS2023 dataset, DWCS-Net achieves a 0.48% improvement over MedNeXt and a 1.34% improvement over nn-UNet, with an HD95 value of 10.61 mm, 0.63 mm lower than MedNeXt. These results confirm that DWCS-Net substantially enhances segmentation accuracy and boundary handling, demonstrating a reasonable performance-efficiency trade-off that supports its suitability for complex vascular structure segmentation.

Background and Objective While the Harris Hawks Optimizer (HHO) is widely utilized for wrapper-based Feature Selection (FS) due to its efficiency and ease of implementation, existing HHO-based FS approaches encounter challenges when handling high-dimensional datasets, such as falling into local optima and high computational costs. In the HHO algorithm, the Harris hawks engage in surprise attacks on the identified prey according to the prey's escape energy. However, there may be scenarios where the prey could escape due to the algorithm's limitations. To enhance the algorithm's prey-capture ability, this article introduces an enhanced HHO algorithm termed Prey Capture Harris Hawks Optimizer (PCHHO). Methods The prey capture strategy incorporates crossover and mutation operators to enhance the algorithm's exploratory-exploitative capabilities. The performance of PCHHO is evaluated on the CEC2017 benchmark suite, where it is compared to HHO, with three enhanced HHO algorithms, nine classical metaheuristic algorithms, and nine improved metaheuristic algorithms. The experimental comparison results are synthesized using the Wilcoxon signed-rank and Friedman tests. Ultimately, a binary form of PCHHO (bPCHHO) is designed for wrapper-based FS and compared with six excellent binary metaheuristics using 15 high-dimensional medical datasets. Results The results demonstrate the excellent performance of the proposed algorithm on the CEC2017 benchmark suite compared to other algorithms, as well as the effectiveness of bPCHHO in evolving a subset of features with 77% reduction in classification error, 8% reduction in computational time, and 73% fewer features selected compared to bHHO. Conclusions The proposed PCHHO and its binary variant bPCHHO exhibit superior performance in both benchmark optimization and wrapper-based FS for high-dimensional medical data, highlighting their potential for practical applications.

The recently proposed Horned Lizard Optimization Algorithm (HLOA) has demonstrated considerable optimization capabilities and the desirable trait of being parameter-free, positioning it as a promising swarm intelligence paradigm. However, its performance is hampered by an imbalance between exploration and exploitation, which tends to result in sluggish convergence rate and early entrapment in local optimal solutions, especially when tackling complex problems with high dimension. To overcome these limitations, this paper introduces ROBEHLOA, an enhanced algorithm that synergistically integrates four strategies: Opposition-Based Learning (OBL) and Biogeography-based Learning Strategy (BLS) to accelerate convergence and augment global search capabilities, complemented by Random Following (RF) and a Hierarchical Structure (HS) to improve performance in complex optimization landscapes. To validate its efficacy in the discrete domain, ROBEHLOA is adapted into a binary variant, designated bROBEHLOA, specifically for the task of Feature Selection (FS). The performance of ROBEHLOA and its binary variant is substantiated through exhaustive experimentation. On the CEC 2017 benchmark functions, ROBEHLOA demonstrates marked superiority over the original HLOA and 19 other well-regarded Metaheuristic algorithms (MAs). In its application to FS, the binary version bROBEHLOA is rigorously evaluated on 20 public datasets, where it consistently achieves a superior balance of higher classification accuracy and more compact feature subsets when compared to nine prominent binary MAs. Collectively, these findings validate ROBEHLOA as a robust and high-performance algorithm for global optimization, establishing its binary variant as an advanced competitive method for FS.

Lupus nephritis (LN) is a serious complication of systemic lupus erythematosus characterized by inflammation and damage to the kidneys, potentially leading to kidney failure. To augment the diagnosis process of LN, innovative minds in the field have intertwined LN imaging with metaheuristic algorithms, thereby effectuating multi-threshold image segmentation (MTIS), and subsequently facilitating a more precise classification of LN. The rime optimization algorithm (RIME) is a physical phenomenon based metaheuristic algorithm. In this work, to enhance RIME's efficiency in MTIS on LN images, the key processes from the partial reinforcement optimizer (PRO) have been integrated into RIME. This integration markedly enhances RIME's convergence precision and speed. In addition, the edge information utilization strategy (EIU) was proposed to improve the diversity of the population. The enhanced RIME (HRIME) has been compared with a series of state-of-the-art algorithms in the IEEE CEC 2017 benchmark functions, demonstrating superior performance. Further testing on three realistic engineering design optimization problems demonstrated the potential of HRIME to address practical challenges. Moreover, the Berkeley segmentation dataset 500 was used to evaluate the efficacy of HRIME for MTIS, and the segmentation method adopted a segmentation technique combining non-local means 2D histogram with Renyi’s entropy. Finally, HRIME was applied to LN image segmentation and the experimental outcomes confirmed that HRIME effectively improved the performance of RIME in LN image MTIS.

[This corrects the article DOI: 10.1016/j.isci.2023.107896.].

Breast cancer is a cancer with a high mortality rate in the medical field. Its pathological images exhibit considerable complexity, characterized by diverse intensity and color patterns in tissue structures and lesion areas. This diversity presents challenges for traditional segmentation techniques in accurately identifying subtle features. In contrast, the multi-threshold image segmentation method optimizes the threshold selection process using meta-heuristic algorithms, enabling more precise segmentation of various tissue types and lesion areas. Therefore, this study introduces a multi-threshold image segmentation model utilizing a modified ant colony optimization algorithm for continuous domains (ACOR), with the goal of enhancing the effectiveness of segmenting breast cancer pathology images. By integrating the Worm Strategy (WS) and Generalized Quadratic Interpolation (GQI) into ACOR, a novel optimization algorithm named GWACOR is developed. Specifically, WS facilitates individuals in jumping out of local optima, thereby bolstering the algorithm’s global search capabilities. GQI strengthens the local exploitation ability of individuals, further leading to enhanced solution accuracy. To evaluate the optimization performance of GWACOR, a series of comparative experiments are carried out on IEEE CEC2017 across three different dimensions. The experimental results show that GWACOR surpasses comparable algorithms in overall performance. Further, a multi-threshold image segmentation model based on GWACOR (GWACOR-MTIS) is formulated by amalgamating GWACOR with the non-local mean two-dimensional histogram and Kapur’s entropy. Through segmentation and comparison with various competing models across different threshold levels, the results demonstrate that GWACOR-MTIS possesses a competitive advantage in segmentation accuracy, providing potentially valuable and dependable diagnostic assistance for clinicians. Consider that the source code of GWACOR is publicly available at https://github.com/EnhancedAlgorithms/GWACOR .

Hyperspectral band selection (BS) aims to extract the most representative features from hundreds of spectral bands, enabling effective dimensionality reduction while preserving interpretability, thereby providing reliable feature support for downstream tasks. However, existing band selection methods based on evolutionary computation hardly balance spectral structure preservation and task adaptability. They often excessive emphasis on improving optimization strategies themselves, while neglecting the structural discrepancies between hyperspectral data characteristics and downstream task requirements. To address these issues, this paper proposes a hyperspectral band selection framework based on an improved grey wolf optimization algorithm (DIGWO_BS). The proposed framework consists of three key components: a multi-layer band preprocessing and encoding mechanism (MBP), a spectral relationship representation method (SSE), and an improved grey wolf optimization algorithm (DIGWO). Specifically, MBP progressively removes noisy and redundant bands prior to optimization, SSE constructs a spectral relationship representation space by integrating spectral information entropy with intra-class and inter-class variations, and DIGWO incorporates cooperative and competitive strategies during the optimization process to enhance search capability. Experimental results on three representative hyperspectral datasets (Indian Pines, Pavia University, and Kennedy Space Center) demonstrate that MBP effectively reduces band redundancy and improves the overall quality of the input spectra, SSE enhances band discriminability while preserving spectral structural consistency, and DIGWO achieves superior classification performance using fewer selected bands through more efficient global search. Overall, DIGWO_BS outperforms several comparative methods in terms of classification accuracy, band redundancy suppression, and spectral structure preservation, providing an efficient, stable, and interpretable solution for hyperspectral band selection.

The Rime Optimization Algorithm (RIME) is a novel metaheuristic approach inspired by the natural formation of rime ice. Despite being a metaheuristic algorithm, RIME still exhibits limitations such as slow convergence speed, low convergence accuracy, and vulnerability to local optima. To overcome its drawbacks, this paper introduces an improved RIME algorithm, EDRIME. The EDRIME algorithm integrates an Exponential Distribution Optimizer (EDO) and a Directional Crossover (DX) strategy to accelerate the convergence rate and enhance global search capabilities. This paper compares EDRIME with 16 advanced algorithms on the CEC2014 and CEC2022 benchmarks to validate its performance in global optimization, including advanced improved algorithms such as EBOwithCMAR and LSHADE_cnEpSi. Experiments have shown that EDRIME outperforms other comparative algorithms in different problem dimensions and testing benchmarks. With the continuous expansion of data scale, the cost and complexity of solving Feature Selection (FS) problems are increasing. To address the challenge of high-dimensional feature selection, this study introduces a novel method named BEDRIME-KELM, which integrates EDRIME with the KELM classifier. The efficacy of the method is validated across 12 diverse datasets sourced from the UCI repository. The results demonstrate that BEDRIME-KELM exhibits strong competitiveness in classification accuracy and feature reduction.

In the field of complex optimization problem solving, the Artemisinin Optimization (AO) simulates the multi-stage synergistic mechanism of artemisinin in the treatment of malaria. It realizes a multi-stage collaborative search through adaptive probability transformation and dynamic weight adjustment and balances global exploration and local exploitation. However, its global exploration adopts fixed probability search, which leads to insufficient exploitation and slow convergence. Static mutation parameters will cause invalid disturbance and destroy the potential solution, so there are still defects such as low convergence accuracy and sub-optimal when solving complex multimodal problems. Therefore, this paper proposes an improved AO, termed GOLAO, which integrates Three-level Synergistic Gaussian Mutation (TSGM) and Adaptive Orthogonal Learning (AOL). TSGM dynamically optimizes the search agent through the three-level coordination mechanism of’state-space-probability’, improves the diversity of the population, and significantly reduces the risk of search agents falling into sub-optimal. AOL reduces redundant computation through systematic exploration, enhances its spatial coverage ability, and improves convergence speed and precision. The performance of the GOLAO algorithm was evaluated using the benchmark function test sets from IEEE CEC2014 and IEEE CEC2022, with comparisons to other peer algorithms. The findings indicate that GOLAO demonstrates strong performance. Furthermore, to assess the practical applicability of the GOLAO algorithm in real-world optimization challenges, it is employed for breast cancer pathology image segmentation. Comparative analyses with alternative approaches are conducted across six distinct threshold levels. The results show that GOLAO has apparent advantages in the image segmentation effect.

Metaheuristic algorithms have become essential for addressing high-dimensional medical optimization problems, yet many still suffer from premature convergence and limited precision. To overcome these issues, we propose a diversity-integrated weighted ranking strategy that combines fitness and population diversity to construct a novel ranking matrix, reshaping convergence dynamics and achieving a better balance between exploration and exploitation. On the IEEE CEC 2017 benchmark suite, this strategy achieves significant improvements on 29 of 30 functions. Building on this foundation, we integrate the mechanism into the Artemisinin Optimization Algorithm and develop WRAO, which is further applied to two key medical tasks: feature selection and medical image segmentation. In feature selection, WRAO effectively identifies compact and discriminative medical features, enhancing disease classification accuracy and interpretability. In image segmentation, it optimizes boundary localization and structural consistency across varying imaging conditions, facilitating more precise pathological analysis. These applications demonstrate that WRAO not only advances metaheuristic optimization but also provides a robust and adaptable computational tool for medical data analysis and decision support, capable of meeting the precision and reliability demands of complex clinical environments.

Population-based optimization algorithms have become increasingly prominent in addressing complex real-world problems; however, they often face criticism due to insufficient performance, lack of rigorous theoretical foundation, and an over-reliance on metaphorical concepts. To overcome these limitations, this paper introduces the LangEvin Equation (LEE) based Evolutionary algorithm, a novel optimization framework grounded in physical laws and population dynamics. LEE employs the mathematical principles of the LangEvin equation, which models the stochastic motion of particles under thermal fluctuations, to guide the search process in both global and local spaces. The algorithm features three innovative operators: the Search Mechanism Operator (SMO), which generates new candidate solutions using LangEvin dynamics to promote effective exploration; the Diversity Promoter (DP), which maintains population diversity and prevents premature convergence by encouraging a broader search; and the Local Escaping Operator (LEO), which enables the algorithm to escape from local optima and accelerates convergence towards high-quality solutions. This synergistic combination endows LEE with a simple yet robust architecture, making it suitable for both constrained and unconstrained optimization tasks. Comprehensive experimental studies were conducted on 23 benchmark functions, the IEEE CEC-2017 and CEC-2022 benchmark suites, as well as six challenging engineering design problems. The results demonstrate that LEE is more successful than well-known existing optimizers in terms of convergence speed, solution accuracy, and stability. As an open-source tool, LEE offers a reliable, efficient, and theoretically grounded approach for tackling optimization challenges in artificial intelligence, machine learning, and engineering domains, providing significant potential for further research and practical applications.

In low-altitude freight logistics scenarios, Unmanned Aerial Vehicle (UAV) path planning plays a critical role in ensuring operational efficiency and collision-free navigation. The Escape algorithm (ESC) represents a promising metaheuristic approach for addressing UAV path planning challenges. However, due to its core limitations of a static fixed-ratio population partitioning strategy and a single search scheme during the late iteration phase, the search behavior of individuals within the population lacks clear guidance and falls into blind exploration. This limits ESC to achieving ideal optimization results only in simple scenarios; when applied to complex UAV navigation environments, its drawback of inadequate solution accuracy will be thoroughly exposed. To mitigate these drawbacks, this study presents an adaptive decentralized search strategy designed to effectively mitigate ESC’s shortcomings. Specifically, the strategy dynamically partitions the population during iterations. Using the average fitness value as a threshold, it divides the population into the elite group and the general group, and assigns differentiated search strategies to each for co-evolution. By integrating this strategy, we develop a novel algorithm referred to as ADSESC. ADSESC is evaluated against other state-of-the-art algorithms on 29 IEEE CEC 2017 benchmark functions and the UAV path planning models based on four constraints. The results demonstrate that ADSESC not only excels in performance across the 29 benchmark functions but also generates smoother and shorter paths compared to the other algorithms evaluated.

Skin cancer is a significant global health concern, with melanoma accounting for approximately 80% of cancer-related deaths. Early diagnosis and treatment are crucial for increasing patient survival rates. Dermoscopy serves as a valuable tool for early detection, where efficient image processing enhances diagnostic accuracy. crucial step in this process is image segmentation, particularly multi-threshold image segmentation (MIS), known for its simplicity, interpretability, and effectiveness. However, conventional meta-heuristic optimizers used improve MIS often become trapped in local optima, reducing their effectiveness. To address this limitation, paper proposes an enhanced RIME optimization (RIME), termed RDRIME, which incorporates a random ferential strategy to mitigate the issue of local optima. The proposed RDRIME-based MIS method was effectively applied to the segmentation of dermoscopic skin cancer images. It utilizes two-dimensional (2D) Renyi's entropy as the objective function and employs a non-local means 2D histogram to comprehensively represent image information. Benchmark tests demonstrate that RDRIME outperforms the original RIME, four improved RIME algorithms, and ten advanced algorithms in avoiding local optima. Experiments on dermoscopic skin cancer images demonstrate that the proposed RDRIME-based MIS method achieves superior segmentation quality greater adaptability across different threshold levels than competing optimization-based methods. In addition, comparisons on the public ISIC-2017 dataset with five representative deep learning models show that RDRIME remains highly competitive in segmentation accuracy without requiring model training, achieving a mean score of 0.9360, only 0.28% below the best-performing network, while exhibiting the lowest standard deviation. These results indicate that RDRIME provides an accurate, stable, and interpretable alternative for skin lesion segmentation.

Chronic obstructive pulmonary disease (COPD) poses a significant threat to global public health, with its diagnosis heavily reliant on accurate segmentation of medical computed tomography (CT) images. However, current segmentation methods suffer from limited optimization efficiency, accuracy, and high computational costs. To address these issues, this paper proposes an improved artemisinin optimization algorithm (ADSPAO), which incorporates spiral propagation (SP) and adaptive delivery (AD) strategies to enhance local exploitation and global exploration, respectively. Experimental comparisons on the IEEE CEC2017 benchmark show that ADSPAO exhibits superior robustness and stability across various function types. Furthermore, the algorithm is applied to multi-threshold segmentation of COPD CT images using Renyi entropy as the fitness function. Results demonstrate that ADSPAO achieves better segmentation performance in terms of peak signal to noise ratio (PSNR), structural similarity (SSIM), and feature similarity (FSIM) compared to other methods, offering an efficient and reliable solution for medical image segmentation.

Metaheuristic algorithms have emerged as indispensable tools for solving NP-hard optimization problems that defy traditional methods. To advance the field’s focus on algorithmic performance, this study introduces the Theory Evolution Optimization (TEO) – an efficient metaheuristic inspired by the evolution of scientific theory. TEO simulates the competitive, accumulative, and replacement processes among scientific hypotheses, mirroring the evolution from a hypothesis to an established scientific theory. The performance of TEO is validated through extensive experimental simulations and benchmarked against 28 popular algorithms, including highly competitive champions such as EBOwithCMAR, LSHADE_cnEpSi, and LSHADE. Pairwise comparisons between TEO and the latest algorithms are conducted using the Wilcoxon signed-rank test, with multiple comparisons managed by the Friedman test. Initially, TEO is tested on the classical IEEE CEC2017 and the latest IEEE CEC2022 benchmark functions. TEO successfully addresses four prominent engineering design problems in constrained continuous space for practical applications. Additionally, a binary TEO (BTEO) variant is introduced and applied to feature selection tasks in discrete space. Experimental results consistently demonstrate that TEO proposes highly competitive outcomes in optimization problems. The source codes for this research are accessible to the public at https://aliasgharheidari.com/TEO.html .

Pulmonary hypertension (PH) is a progressive cardiopulmonary disorder with high mortality, necessitating non-invasive methods for early detection and treatment evaluation. In this paper, this study proposes a novel machine learning model for non-invasively identifying the therapeutic effects of Baicalin in PH using routine hematological indicators. The core innovation is an enhanced Bat Algorithm (BA) variant, termed RGBA, which integrates an Elite-based Random Walk Strategy (ERWS) and an Elite Guided Strategy (EGS) to achieve a superior balance between exploration and exploitation. And the RGBA demonstrated significantly improved global search capability in IEEE CEC 2014 benchmark tests, outperforming several state-of-the-art meta-heuristic algorithms. Subsequently, a binary version of RGBA (bRGBA) was developed and combined with a Kernel Extreme Learning Machine (KELM) classifier within a wrapper-based feature selection framework, forming the bRGBA-KELM model. Applied to a dedicated PH dataset from murine models, bRGBA-KELM achieved a prediction accuracy of 97.43% via ten-fold cross-validation, outperforming nine comparable hybrid models. Critically, it identified four key blood biomarkers-Red Blood Cell count (RBC), Hemoglobin (HGB), Mean Corpuscular Volume (MCV), and Hematocrit (HCT)-that are mechanistically linked to PH pathogenesis and modulated by Baicalin treatment. In conclusion, the proposed RGBA offers a robust optimization tool, while the bRGBA-KELM model provides a clinically viable, non-invasive technical reference for early PH prediction and therapeutic assessment.