Bone fractures in dogs are common orthopaedic conditions that require accurate diagnosis and rapid intervention. Traditional radiographic interpretation is often time consuming and is subject to variability, emphasizing the need for automated diagnostic tools. This paper represents a deep learning model based on classification of long bone fractures in dogs using medical conventional radiographic images. The proposed model uses a convolutional neural network (CNN), specifically ResNet50 to improve detection and fracture classification. Comparative analysis with other deep learning architectures, including VGG16 and MobileNeTV2, shows the excellent ResNet50 performance. To address the challenge of limited annotated veterinary radiographic datasets, the actual data strategy of augmentation is implemented, which increases the generalization of the model. In addition, the segment model of anything (SAM) is integrated for automated fracture segmentation, allowing precise location and improving diagnostic efficiency. Experimental results show that the ResNet50 achieves high classification performance with an accuracy of 99.76%, accuracy of 99.53%, 100% and F1-score 99.76%, overcoming other competing architecture. These findings emphasize the potential of artificial intelligence in veterinary orthopaedic and offer an efficient, accurate and automated solution for diagnosis and control of long bone fractures in social animals.

The rising global incidence of Multiple Sclerosis (MS), an autoimmune disorder that impacts the central nervous system, demands novel diagnostic approaches for early detection and intervention. Given the challenges of irreversible MS progression and the complexity of traditional diagnosis methods, this study introduces a hybrid Artificial Intelligence (AI) framework that enhances MS diagnosis accuracy using MRI scans. Our model integrates a multi-view ResNet architecture with novel attention mechanisms—View Space Attention Block (VSAB) and View Channel Attention Block (VCAB)—to extract detailed features from 2D brain images. Additionally, we developed the Quantum RIME (QRIME) algorithm, which combines RIME and Quantum Behaved Particle Swarm Optimization (QPSO) for efficient dimensionality reduction, optimizing both accuracy and computational efficiency. The model was rigorously evaluated using sixteen UCI benchmark datasets and a dedicated brain MRI dataset of 425 scans (262 MS patients and 163 healthy controls), achieving a notable accuracy of 98.29%, precision of 96.49%, specificity of 97.65%, and an F1-score of 97.85%. These results not only demonstrate our model's exceptional capability in identifying MS with high precision but also highlight its potential applicability in diagnosing other neurological disorders. By emphasizing the transformative potential of AI in medical diagnostics, our work underlines the significance of innovative AI applications in enhancing early detection, ultimately aiming to enhance patient outcomes in the neurodegenerative disease domain.

Gender classification from panoramic dental X-ray images has significant applications in forensic identification and clinical dentistry. This study proposes a novel hybrid approach combining deep learning with metaheuristic optimization for accurate gender prediction. Our primary model employs a fine-tuned DenseNet121 architecture for robust feature extraction from dental radiographs, enhanced by a two-stage feature selection process using Particle Swarm Optimization (PSO) and Grey Wolf Optimizer (GWO). For comprehensive evaluation, we compare this against two alternative implementations: a custom Convolutional Neural Network designed specifically for dental images and a fine-tuned ResNet50 model. The custom CNN processes grayscale images and learns features directly from the dataset, while both DenseNet121 and ResNet50 utilize RGB inputs with transfer learning from pre-trained weights. All extracted features undergo optimization through our hybrid PSO-GWO algorithm before classification with Logistic Regression, ensuring both computational efficiency and model interpretability. Experimental results demonstrate that our DenseNet-based approach achieves superior performance with 95.88 % accuracy, outperforming both the ResNet50 model (93.81%) and custom CNN (91.75%). This performance advantage, establishes the effectiveness of combining deep feature extraction with metaheuristic optimization for dental image analysis. The study provides a practical framework for developing accurate, efficient diagnostic tools in dental forensics and clinical practice.

This article introduces a novel optimization approach to improve the parameter estimation of proton exchange membrane fuel cells (PEMFCs), which are critical for diverse applications but are challenging to model due to their nonlinear behavior. The proposed method, HGS-MPA, enhances the Hunger Games Search (HGS) algorithm by integrating Marine Predator Algorithm (MPA) operators, significantly boosting its exploitation capabilities and convergence rate. The effectiveness of HGS-MPA was validated on three commercial PEMFC datasets: 250-W stack, BCS 500-W, and NedStack PS6, using the Sum Squared Error (SSE) as the performance metric. Experimental results highlight that HGS-MPA achieves minimum fitness values of 0.33770, 1.31620, and 0.01174 for the respective datasets, outperforming other state-of-the-art algorithms. These findings underscore the method’s potential for accurate PEMFC parameter estimation, offering enhanced performance and reliability.

The growing adoption of Android-based Internet of Medical Things (IoMT) devices has unfortunately attracted the attention of cyber attackers who exploit vulnerabilities through malware infiltration, resulting in serious threats such as data theft and unauthorized access. However, directly sharing users’ raw data for centralized model training raises privacy concerns. To address these challenges, this paper proposes a federated deep learning-based framework for malware detection and data protection in edge-enabled intelligent healthcare systems. This framework utilizes a distributed deep neural network that is trained without exchanging user row data, ensuring high detection accuracy while preserving user privacy. Two real-world datasets are employed to evaluate the efficiency of the proposed approach, and a comparative analysis is conducted against a baseline approach across three test scenarios. Experimental results demonstrate the superiority of the proposed framework over the baseline techniques, achieving an impressive accuracy of 98.84

This article presents a photovoltaic (PV) optimization approach using the puma optimizer (PO) approach, which has been enhanced by utilizing Lévy flight optimization. The name of this approach is modified puma optimizer (MPO). PV generation systems are essential for sustainable solar energy utilization. It is an innovation and clean energy. There is an urgent demand for suitable and reliable simulation and optimization techniques for PV systems. This will result in increased efficiency. Algorithms with a high degree of reliability are needed to ensure optimal PV parameters. This study was conducted with MATLAB software. This article introduces the original PO method as a means to evaluate the performance of the MPO approach. The root mean square error (RMSE) function serves as a benchmark. Based on the simulation findings, the MPO approach shows superior RMSE compared to the PO method, specifically at a value of 0.0026%.

Yield augmentation of solar distiller (SD) using low-cost modifications with minimum environmental impacts is highly desirable. In this study, we developed an eco-friendly steam generation device made of carbonized Ficus sycomorus wood to boost the yield of the conventional single-basin solar distiller. Moreover, a hybrid model composed of a long short-term memory (LSTM) network and a Starling murmuration optimizer (SMO) was developed to predict the yield of the established SD. To verify the accuracy of the model, three other models were employed and the distillate yield was predicted. The predicted data of all models was evaluated using different error measures. These models were standalone LSTM and LSTM optimized by an artificial hummingbird algorithm (AHA) or manta ray foraging optimizer (MRFO). Real field data were employed during the training and testing of all models. The thermal performance of the distiller with the steam generation device was compared with that of a traditional distiller based on exergy output, exergy efficiency, and energy efficiency. The distillate yield, exergy efficiency, and energy efficiency of the modified SD were enhanced by 34 %, 49 %, and 40 %, respectively, compared with that of the traditional distiller. Moreover, the total amount of produced drinkable water using the modified SD could reach a high value of 6.1 L/day with a low cost of 0.014 $/L.

This work proposes a novel Transformer based on the Kolmogorov–Arnold Network (TFKAN) model for Intrusion Detection Systems (IDS) in the IoT environment. The TFKAN Transformer is developed by implementing the Kolmogorov–Arnold Networks (KANs) layers instead of the Multi-Layer Perceptrons (MLP) layers. Unlike the MLPs feed-forward layer, KAN layers have no fixed weights but use learnable univariate function components, enabling a more compact representation. This means a KAN can achieve comparable performance with fewer trainable parameters than a larger MLP. The RT-IoT2022, IoT23, and CICIoT2023 datasets were used in the evaluation process. The proposed TFKAN Transformer outperforms and obtains higher accuracy scores of 99.96%, 98.43%, and 99.27% on the RT-IoT2022, IoT23, and CICIoT2023 datasets, respectively. The results indicate that the developed Transformer using KAN shows promising performance in IDS within IoT environments compared to MLP layers.Transformers based on KANs are on average 78% lighter, in parameter count, than Transformers using MLPs. This makes KANs promising to be a replacement for MLPs.

In this study, a tubular solar air heater with swirl-flow and find absorber supported by radial and longitudinal fins (SAH) was used to improve thermohydraulic performance. The suggested SAH has been compared to the plain duct SAH without fins at a typical air flow rate of 0.01-0.050 kg/s. The results of this study show that radial fins improve SAH performance characteristics when compared to a plain heater or a heater with longitudinal fins. Furthermore, increasing the number of radial fins and airflow rate improves SAH performance. SAH with five radial fins and an airflow rate of 0.050 kg/s achieves maximum thermal and thermohydraulic performance. The experimental analysis showed that the pressure loss in the proposed SAH with swirl tubes increases by 2.64 % and 16.44 % for AMFR 0.010 kg/s and 0.030 kg/s, respectively. Additionally, radial fins record the highestpressure loss values, with an increase in pressure loss of almost 11.97 % with maximum average temperature difference about 9.75 degrees C. Also, this study also demonstrates the effectiveness of various Kolmogorov-Arnold Networks (KAN) models combined with optimization algorithms in predicting the outlet temperature and pressure loss of the investigated heater. The analysis, based on experimental field data, uses five optimization techniques: KAN-CGO (Chaos Game Optimization), KAN-LASHDE (Lagrangian Adaptive Sheepdog Differential Evolution), KAN-AOA (Arithmetic Optimization Algorithm), KAN-HHO (Harris Hawks Optimizer), and KAN-WO (Walrus Optimizer). The KAN-WO model achieved an RMSE of 0.7557, demonstrating that its predictions closely match the experimental data. So, the KAN-WO model emerges as the most reliable for predicting outlet temperature and pressure loss, due to its superior performance, consistency, stability, and ability to generalize.

Seawater intrusion (SWI) in coastal aquifers in arid areas is a global threat to the groundwater quality. This study aimed to assess SWI in the Sudr coastal aquifer in South Sinai, Egypt, by integration of hydrochemical, isotopic, remote sensing (RS) and machine learning (ML) approaches. The developed ML technique is based on the random vector functional link (RVFL) using the Tactical unit algorithm (TAU) to predict the value of the parameters of the RVFL, which leads to an enhancement in the prediction of seawater intrusion indicators. The hydrogeochemical data of the groundwater wells in the two periods (1996 and 2022) were used as inputs for the ML method. The groundwater chemistry in the study area was in the following order: Na > Ca > Mg > K and Cl > SO4 > HCO3. Ionic ratios highlighted groundwater–seawater mixing and reverse ion exchange processes. Based on the Cl−/Br− ratio and O18 D data, seawater intrusion is the primary source of groundwater salinization. The results show that 83.33

Introduction:Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by challenges in communication, social interactions, and repetitive behaviors. The heterogeneity of symptoms across individuals complicates diagnosis. Neuroimaging techniques, particularly resting-state functional MRI (rs-fMRI), have shown potential for identifying neural signatures of ASD, though challenges such as high dimensionality, noise, and small sample sizes hinder their clinical application. Methods:This study proposes a novel approach for ASD detection utilizing deep learning and advanced feature selection techniques. A hybrid model combining Stacked Sparse Denoising Autoencoder (SSDAE) and Multi-Layer Perceptron (MLP) is employed to extract relevant features from rs-fMRI data in the ABIDE I dataset, which was preprocessed using the CPAC pipeline. Feature selection is enhanced through an optimized Hiking Optimization Algorithm (HOA) that integrates DynamicOpposites Learning (DOL) and Double Attractors to improve convergence toward the optimal subset of features. Results:The proposed model is evaluated using multiple ASD datasets. The performance metrics include an average accuracy of 0.735, sensitivity of 0.765, and specificity of 0.752, surpassing the results of existing state-of-the-art methods. Discussion:The findings demonstrate the effectiveness of the hybrid deep learning approach for ASD detection. The enhanced feature selection process, coupled with the hybrid model, addresses limitations in current neuroimaging analyses and offers a promising direction for more accurate and clinically applicable ASD detection models.

This paper presents a novel optimization algorithm called the Phototropic Growth Algorithm (PGA), which takes its inspiration from the growth patterns of plant cells in response to sunlight. The proposed algorithm has been evaluated on two benchmark test function suites, the CEC 2017 and CEC 2020, and compared to other established metaheuristic-optimization algorithms based on the quality of the solutions obtained. The effectiveness of PGA is further demonstrated by testing it on six challenging engineering design problems with constraints, showcasing its robustness in solving real-world problems with unknown search spaces. Simulation results show that PGA outperforms other competitive algorithms by consistently generating superior solutions and demonstrates strong potential for solving complex optimization tasks using a nature-inspired approach.

This paper provides a comprehensive comparative analysis of deep learning and transformer-based models of the state of the art military aircraft detection from aerial imagery. We evaluate the performance of some models like EfficientNetB3, EfficientNetB5, Vision Transformer (ViT), and Swin Transformer on a specialized military aircraft dataset. Our experimental results demonstrate EfficientNetB3 to be 93 % accurate in the classification of military aircraft with computational efficiency, according to the compound scaling principles established by Tan and Le. We also employ RT-DETR for object detection, which is $92.7 \% \text{mAP}$ and 90.4 % recall on 81 different types of military aircraft. For comparison, YOLOv8 yielded $94 \% \text{mAP}$ and 88.1% recall, and YOLOv7 yielded 90.2% mAP and 82.7 % recall, pushing the real-time detection benchmark established by Redmon et al. further. We use aggressive data augmentation techniques and preprocessing pipelines for enhancing model generalizability.

Time-series forecasting, prediction, and estimation have been active research topics in recent years due to their critical role in decision-making and future planning. Artificial neural networks (ANNs) are widely employed for forecasting applications and have demonstrated significant performance improvements. In this paper, we utilize a specialized type of ANN, the dendritic neuron model (DNM), for time-series forecasting. We apply an optimized DNM model by utilizing the power of metaheuristic (MH) optimization algorithms. Specifically, we employ an MH algorithm, the mountain gazelle optimizer (MGO), to optimize the parameters of the DNM, improve its configuration process, and ultimately improve the accuracy of the forecast. Furthermore, we compare MGO performance with several other optimization algorithms used to optimize the DNM model. Comprehensive evaluations are conducted on four publicly available time series datasets using multiple evaluation indicators. The results demonstrate the superiority of the MGO-DNM model over competing methods, with optimized DNM configurations consistently outperforming the traditional DNM in forecasting tasks.

Wind power forecasting plays a crucial role in optimizing the integration of wind energy into the grid by predicting wind patterns and energy output. This enhances the efficiency and reliability of renewable energy systems. Forecasting approaches inform energy management strategies, reduce reliance on fossil fuels, and support the broader transition to sustainable energy solutions. The primary goal of this study is to introduce an effective methodology for estimating wind power through temporal data analysis. This research advances an optimized Multilayer Perceptron (MLP) model using recently proposed metaheuristic optimization algorithms, namely the Fire Hawk Optimizer (FHO) and the Non-Monopolize Search (NO). A modified version of FHO, termed FHONO, is developed by integrating NO as a local search mechanism to enhance the exploration capability and address the shortcomings of the original FHO. The developed FHONO is then employed to optimize the MLP for enhanced wind power prediction. The effectiveness of the proposed FHONO-MLP model is validated using renowned datasets from wind turbines in France. The results of the comparative analysis between FHONO-MLP, conventional MLP, and other optimized versions of MLP show that FHONO-MLP outperforms the others, achieving an average Root Mean Square Error (RMSE) of 0.105, Mean Absolute Error (MAE) of 0.082, and Coefficient of Determination (R2) of 0.967 across all datasets. These findings underscore the significant enhancement in predictive accuracy provided by FHONO and demonstrate its effectiveness in improving wind power forecasting.

The process of early diagnosis of polyps is considered a critical point of preventive healthcare, as it can significantly improve the prognosis and treatment outcomes of patients with colorectal cancer. Since the Polyps are constructed in the colon, they can grow up to be cancerous over time. We present an alternative Polyps image segmentation approach according to multilevel thresholding techniques (MLTs) to ensure effective early polyp diagnosis. The developed MLT Polyps image segmentation method depends on improving the performance of the Dung beetle optimizer (DBO) algorithm based on the operators of Triangular Mutation (TMO), Comprehensive learning (CL), and dynamic update of the search domain. We conducted a comprehensive evaluation of the performance of the DCTDBO using eight Polyps images and compared it with other image segmentation methods, ensuring a rigorous and thorough process. The results show the high ability of the DCTDBO according to performance measures to segment the Polyps images. In terms of peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and feature similarity (FSIM), DCTDBO performed better than the basic version of the DBO and other methods. For example, the average of DCTDBO overall threshold levels and tested images in terms of FSIM, SSIM, and PSNR is 0.9668, 0.99217, and 28.9338, respectively. This indicates the influence of TMO and CL on enhancing the performance of DBO.

In this study, we introduce a framework for medical image classification that combines deep learning models with modified Beluga Whale Optimization as a feature selection technique. Our approach utilizes a feature extraction technique called CrossViT, a vision transformer model architecture design of ViT. The CrossViT is used to extract relevant features from medical images. A modified version of the Beluga Whale Optimization (BWO) method is also employed to select the relevant feature. The modified BWO incorporates the Triangular Mutation Operator (TMO) approach to enhance the BWO’s ability to exploit the problem space. A set of twelve functions from the CEC2022 benchmark is used to evaluate the Modified BWO (MBWO) performance and compare it with traditional BWO, followed by assessing the proposed medical image classification framework on several benchmark datasets, demonstrating excellent performance results. This study presents an innovative and effective approach to medical image classification by combining the strengths of deep learning and metaheuristic optimization methods.

The accurate prediction of crude oil production is crucial for effective management of oil reservoir operations. This paper leverages recent advancements in machine learning techniques and metaheuristic optimization algorithms, specifically deep learning (DL) and metaheuristic (MH) approaches, to construct a robust and efficient oil production prediction model. Real-world datasets from two diverse countries, Yemen and China, are employed in model development. The study focuses on optimizing a multilayer perceptron (MLP) using the Runge–Kutta optimizer (RUN). The primary goal is to enhance the MLP parameters through the application of the RUN algorithm. Rigorous evaluation experiments gauge the efficacy of the resulting prediction model (RUN-MLP), demonstrating impressive performance across three widely recognized evaluation metrics: root-mean-square error (RMSE), mean absolute error (MAE), and coefficient of determination ( R^2 ). Comparative analyses involve multiple MLP-modified models employing various MH algorithms, with the RUN-MLP consistently exhibiting competitive performance. The findings underscore the computational efficiency of the RUN optimization algorithm. Additionally, the study employs the Friedman test as a statistical tool to elucidate differences between RUN and its competitors.

This paper presents a new data clustering technique aimed at enhancing the performance of the trainable path-cost algorithm and reducing the computational complexity of data clustering models. The proposed method facilitates the discovery of natural groupings and behaviours, which is crucial for effective coordination in complex environments. It identifies natural groupings within a set of features and detects the best clusters with similar behaviour in the data, overcoming the limitations of traditional state-of-the-art methods. The algorithm utilises a density peak clustering method to determine cluster centers and then extracts features from paths passing through these peak points (centers). These features are used to train the support vector machine (SVM) to predict the labels of other points. The proposed algorithm is enhanced using two key concepts: first, it employs Q-Generalised Extreme Value (Q-GEV) under power normalisation instead of traditional generalised extreme value distributions, thereby increasing modelling flexibility; second, it utilises the random vector functional link (RVFL) network rather than the SVM, which helps avoid overfitting and improves label prediction accuracy. The effectiveness of the proposed clustering algorithm is evaluated through various experiments, including those on UCI benchmark datasets and real-world data, demonstrating significant improvements across multiple performance metrics, including F1 measure, Jaccard index, purity, and accuracy, highlighting its capability in accurately identifying paths between similar clusters. Its average F1 measure, Jaccard index, purity, and accuracy is measured 76.87%, 56.29%, 80.29%, and 79.64%, respectively.

Human Activity Recognition (HAR) is crucial for biomedical signal processing as it provides contextual information that enhances the interpretation and accuracy of physiological data. By identifying and classifying human activities, HAR enables the differentiation between normal and abnormal physiological states. This research addresses key challenges in HAR, including the lack of data diversity, the presence of noise, and imbalanced datasets, which prevent the development of robust and accurate HAR models. To overcome these challenges, we propose a novel ensemble framework, MK-WaveNet, designed to improve the generalization and accuracy of HAR systems. Our framework integrates a model-based data augmentation backbone, termed the Inverse Wavelet Augmented Convolutional Network (IWACN), to enhance feature representation and mitigate data imbalance issues. Additionally, a novel Minkowski distance-based temporal fusion algorithm is introduced to effectively merge the outputs of ensemble models, enhancing the overall performance of the framework. Comprehensive experiments conducted on four benchmark datasets (UCI-HAR, PAMAP2, DAPHNET, and MobiAct) and different backbones demonstrate the efficacy of the proposed framework in addressing the challenges of HAR. MK-WaveNet achieved F1-scores of 98.32% on UCI-HAR, 92.21% on PAMAP2, 94.83% on DAPHNET, and 98.86% on MobiAct. The results demonstrate that our proposed framework, alongside its backbones, sets a new benchmark by outperforming existing state-of-the-art models across all considered datasets.