Temporal Lobe Epilepsy (TLE), a very common type of focal epilepsy, requires early and precise identification to enable prompt treatment and enhance patient outcomes. Traditional methods of detection, most notably manual MRI interpretation, tend to be time-consuming, subjective, and dependent on expert specialists, hindering their scalability and reliability. Due to the limitations of conventional and current automated systems—e.g., poor feature extraction, poor segmentation, and low accuracy in classification—this work presents NeuroTLEDNet, a new end-to-end artificial intelligence model for early TLE detection from medical images. The new methodology integrates multiple series of innovations across every processing phase. In preprocessing, skull stripping, Gaussian blurring, and color normalization are used to improve image uniformity and suppress irrelevant structures. In segmentation, the improved TLE-SegNet + is suggested—based on Mask R-CNN and supplemented with DarkNet-53, DenseNet, HetConv layers, and Attention Feature Pyramid Networks (FPN)—to facilitate highly accurate delineation of epileptogenic areas. For further describing the segmented areas, a rich feature space encompassing color histograms, Local Binary Patterns (LBP), and geometric features (eccentricity, area, perimeter) is developed. To resolve inefficiencies in feature selection, this paper presents the Squirrel Reptilian Optimization (SRO) algorithm, a new hybrid metaheuristic that incorporates both the Squirrel Search Algorithm (SSA) and the Reptile Search Algorithm (RSA). The hybrid enhances discriminative feature selection by handling exploration and exploitation dynamics better compared to individual methods. For categorization, NeuroTLEDNet utilizes an in-depth hybrid ensemble of 3D Convolutional Neural Networks (3D-CNN), Recurrent Neural Networks (RNN), Radial Basis Function Networks (RBFN), and SqueezeNet, facilitating powerful modeling of spatial-temporal relationships in brain MRI scans. Tested on an 80/30 split MRI dataset, the designed model attains a remarkable accuracy of 98.6%, outperforming current methods like CNN-LSTM (94.2%), Mask R-CNN (92.7%), and ResNet-based architectures (93.1%). Indeed, NeuroTLEDNet provides an end-to-end solution for the shortcomings of existing TLE detection methods by utilizing advanced preprocessing, segmentation, feature optimization, and hybrid classification, with state-of-the-art performance in terms of both accuracy and explainability. The suggested model is a promising candidate for clinical application, and it has the potential to assist neurologists in faster and more accurate TLE diagnosis.
Advanced Health Science and Technology Journal · مجلد 2 (1) · 63-72
Abstract Background: Well over 65 million people throughout the world suffered from epilepsy, a common neurological illness. The success rate of standard medical interventions, including medication and surgery, is <30%. On the other hand, seizure prevention methods involving prompt therapeutic interventions are possible with early seizure prediction. Effective and highly accurate epilepsy prediction is still a challenge, despite experts’ best efforts. Aims and Objectives: We present the Particle Swarm Optimized–Convolutional Neural Network and Long Short Term Memory (PSO CNN LSTM) model for electroencephalogram (EEG) signal categorization to support epilepsy diagnosis. Materials and Methods: For EEG classification, we acquire EEG samples from the Bonn dataset that has been used in many reliable studies, which contains five categories of EEG signals, including both normal and epileptic samples. The dataset is preprocessed by removing artifacts and noise. The Short Time Fourier Transform method is subsequently employed for converting the EEG signals into spectrogram images. The postulated PSO CNN LSTM model and other deep learning models, such as CNN and CNN LSTM, employed these spectrogram images as inputs. MATLAB software was implemented for training and verifying models. Results and Conclusions: The studies’ findings revealed that the suggested PSO CNN LSTM model performs better than the other models, with an accuracy of 96% as opposed to 84% and 94%, respectively.
Prompt and accurate classification of wheat leaf diseases and their severity is essential for precise diagnosis, effective pesticide application, efficient disease control, and enhanced wheat production and quality. Nevertheless, the wide range of wheat diseases poses a significant challenge in terms of their identification, especially in intricate agricultural landscapes. The utilization of conventional models has major limitations in wheat disease detection, including dataset-specific performance, overfitting due to limited data, and high computational needs, making deployment in resource-constrained situations difficult. To address these challenges, this research proposes a multi-stage Convolutional Neural Network (CNN) based Ensemble Learning (EL) approach, which utilizes several concurrent CNN models with a bagged EL method to classify wheat leaf diseases. The various stages in the EL approach employ a multi-level framework to enhance feature extraction and capture complex data patterns. This improves the model’s emphasis on diseases while reducing the influence of complex backgrounds on disease identification. The proposed method, integrating pretrained CNNs and a bagging ensemble technique, achieved an accuracy of 86.78%, 98.28%, and 99.16% on the three publicly available datasets, outperforming the state-of-the-art models. These results demonstrate the potential of the proposed model for real-time disease diagnosis.
In the present scenario, cancerous tumours are common in humans due to major changes in nearby environments. Skin cancer is a considerable disease detected among people. This cancer is the uncontrolled evolution of atypical skin cells. It occurs when DNA injury to skin cells, or a genetic defect, leads to an increase quickly and establishes malignant tumors. However, in rare instances, many types of skin cancer occur from DNA changes tempted by infrared light affecting skin cells. This disease is a worldwide health problem, so an accurate and appropriate diagnosis is needed for efficient treatment. Current developments in medical technology, like smart recognition and analysis utilizing machine learning (ML) and deep learning (DL) techniques, have transformed the analysis and treatment of these conditions. These approaches will be highly effective for the recognition of skin cancer utilizing biomedical imaging. This study develops a Multi-scale Feature Fusion of Deep Convolutional Neural Networks on Cancerous Tumor Detection and Classification (MFFDCNN-CTDC) model. The main aim of the MFFDCNN-CTDC model is to detect and classify cancerous tumours using biomedical imaging. To eliminate unwanted noise, the MFFDCNN-CTDC method initially utilizes a sobel filter (SF) for the image preprocessing stage. For the segmentation process, Unet3+ is employed, providing precise localization of tumour regions. Next, the MFFDCNN-CTDC model incorporates multi-scale feature fusion by combining ResNet50 and EfficientNet architectures, capitalizing on their complementary strengths in feature extraction from varying depths and scales of the input images. The convolutional autoencoder (CAE) model is utilized for the classification method. Finally, the parameter tuning process is performed through a hybrid fireworks whale optimization algorithm (FWWOA) to enhance the classification performance of the CAE model. A wide range of experiments is performed to authorize the performance of the MFFDCNN-CTDC approach. The experimental validation of the MFFDCNN-CTDC approach exhibited a superior accuracy value of 98.78% and 99.02% over existing techniques under ISIC 2017 and HAM10000 datasets.
skin cancer detectionskin lesion classificationmulti-scale feature fusionconvolutional neural networks
Congenital heart disease (CHD), impacting around 1 % of infants worldwide, constitutes a significant healthcare challenge. Early detection is crucial, however constrained by the intricacies of conventional diagnostic techniques such as auscultation and echocardiography. This research presents a tailored one-dimensional convolutional neural network (1D-CNN) for the classification of phonocardiogram (PCG) signals into normal or abnormal categories, providing an automated and efficient solution for congenital heart disease (CHD) diagnosis. The model was trained on a composite dataset consisting of local pediatric PCG signals and publicly accessible dataset. Preprocessing methods, such as low- and high-pass filtering (60-650 Hz), resampling, and noise reduction, were utilized to enhance signal quality. Data augmentation techniques, including chunking, padding, and pitch-shifting, were employed to rectify dataset imbalances and improve model efficacy. Experimental results indicate substantial enhancements, attaining an accuracy of 98.56 %, precision of 98.56 %, F1 score of 98.55 %, sensitivity of 0.98, and specificity of 0.99. The comparative analysis demonstrates the proposed approach's superiority over current methods regarding accuracy and reliability. The research highlights the promise of combining modern signal processing with deep learning for efficient CHD screening. The suggested model exhibits outstanding performance yet, issues like dataset variability and noise persist. Future endeavors involve extending to multiclass categorization and assessing performance across a wider range of medical problems. This study represents a significant advancement in accessible, automated CHD diagnoses, enhancing clinical competence to elevate pediatric treatment.
Breast cancer remains the most prevalent cancer among women, where accurate and interpretable analysis of pathology images is vital for early diagnosis and personalized treatment planning. However, conventional single-network models fall short in balancing both performance and explainability—Convolutional Neural Networks (CNNs) lack the capacity to capture global contextual information, while Transformers are limited in modeling fine-grained local details. To overcome these challenges and contribute to the advancement of Explainable AI (XAI) in precision cancer diagnosis, this paper proposes MVT-OFML (Multi-View Transformer Online Fusion Mutual Learning), a novel and interpretable classification framework for breast cancer pathology images. MVT-OFML combines ResNet-50 for extracting detailed local features and a multi-view Transformer encoding module for capturing comprehensive global context across multiple perspectives. A key innovation is the Online Fusion Mutual Learning (OFML) mechanism, which enables bidirectional knowledge sharing between the CNN and Transformer branches by aligning both intermediate feature representations and prediction logits. This mutual learning framework enhances performance while also producing interpretable attention maps and feature-level visualizations that reveal the decision-making process of the model—promoting transparency, trust, and clinical usability. Extensive experiments on the BreakHis and BACH datasets demonstrate that MVT-OFML significantly outperforms the strongest baseline models, achieving accuracy improvements of 0.90% and 2.26%, and F 1 -score gains of 4.75% and 3.21%, respectively. By integrating complementary modeling paradigms with explainable learning strategies, MVT-OFML offers a promising AI solution for precise and interpretable breast cancer diagnosis and prognosis, supporting informed decision-making in clinical settings.
breast cancerhistopathological image classificationexplainable AICNN-Transformer hybrid
Background As a crucial cognitive and motor function, handwriting can reveal dyslexia, dysgraphia, and other neurodevelopmental conditions. With the emergence of AI, deep learning (DL) models can analyze handwriting patterns for early identification and categorization of these conditions with high level of accuracy. Objective This study presents a literature review by examining DL models in classifying learning disabilities using handwriting data. It identifies models’ shortcomings and suggests future research directions, improving model accuracy, fairness, and usability. Methodology By employing the standard literature guidelines, the authors identified 24 studies from a pool of 502 studies. They extracted dataset details, methodologies, and key findings from the studies. Findings The findings indicate the dominance of convolutional neural networks in improving learning disabilities classification through the extraction of fine-grained spatial handwritten patterns. Vision transformers (ViTs) show potential for long-range contextual feature extraction to enhance the model’s generalization capabilities. Despite these advances, dataset size, model generalizability, computational efficiency, interpretability, and real-world deployment continue to be significant challenges. Conclusions The review findings underscore the transformative potential of DL-based handwriting analysis in educational and clinical settings, providing early detection and intervention in learning disabilities.
Journal of Pioneering Medical Science · مجلد 14 (09) · 103-107
Coronary Artery Disease (CAD) remains a leading cause of morbidity and mortality worldwide. To evaluate the diagnostic performance of AI models in the early detection and risk stratification of CAD using ECG and imaging data and to compare their accuracy with cardiologist interpretation. Methods: This cross-sectional analytical study was conducted in Saudia, included 185 patients with suspected CAD who underwent both ECG and imaging evaluation. AI-based models, including convolutional neural networks for ECG and deep learning algorithms for imaging, were applied to detect CAD and stratify patients into risk categories. Cardiologist-confirmed diagnosis served as the reference standard. Results: The AI-ECG model achieved a sensitivity of 88.5%, specificity of 82.0% and an AUC of 0.90 (95% CI: 0.86–0.94). When ECG and imaging data were combined, diagnostic accuracy improved, with sensitivity of 92.4%, specificity of 85.2% and an AUC of 0.93 (95% CI: 0.89–0.96). AI-based risk stratification categorized 54 patients (29.2%) as low risk, 78 (42.2%) as intermediate risk and 53 (28.6%) as high risk. Confirmed CAD prevalence correlated strongly with AI-predicted risk groups, with 22.2% in the low-risk group and 88.7% in the high-risk group. Conclusion: Artificial intelligence demonstrates high accuracy in the early detection and risk stratification of CAD using ECG and imaging data. AI models performed comparably to cardiologists and offered significant efficiency gains. Integration of AI into cardiovascular workflows may enable earlier intervention, optimized resource allocation and improved patient outcomes. Further validation across larger and more diverse populations is warranted.
Across the globe, chronic obstructive pulmonary disease (COPD) continues to be a critical healthcare burden due to its progressive and undetected decline in pulmonary function. In recent times, convolutional neural networks (CNNs) and vision transformers (ViTs) have been widely applied to lung sounds (LS) analysis, leading to the development of improved non-invasive, accessible, and scalable COPD screening methods. However, the shortcomings of CNNs and ViTs architectures restrict the COPD classifier’s discriminative power in distinguishing normal and COPD samples. In this study, a unique and interpretable COPD classification is proposed. This study introduces a hybrid feature extraction strategy integrating CNNs and ViTs architecture to capture the crucial COPD patterns from the LS’s Mel-spectrogram representation. The proposed CNNs-ViTs architecture addresses the limitations of standalone CNNs or ViTs architectures by capturing fine-grained acoustic features and long-range temporal dependencies while efficiently modeling non-linear relationships. A novel feature fusion is used to combine the heterogeneous features into a unified and highly discriminative feature representation space. A tailored hyperparameter optimization strategy is developed to fine-tune the Kolmogorov-Arnold Networks (KANs) for classifying the features into normal and COPD classes. By incorporating internal (552 Healthy and 1324 COPD LS recordings) and external (35 Healthy and 77 COPD LS recordings) datasets, the authors evaluate the model’s performance. The model achieves a generalization accuracy of 97.62% and specificity of 96.51% on the external dataset. Through Shapley Additive Explanations (SHAP) overlays, the model’s interpretability is enhanced. The experimental findings highlight the effectiveness of the proposed COPD classification architecture in the domain of biomedical acoustic signal processing. Operating with minimal reliance on specialized diagnostic infrastructure renders the proposed model ideal for primary care and resource-limited environments.
Advanced Health Science and Technology Journal · مجلد 1 (2) · 92-99
Abstract Background: Epilepsy remains one of the most prevalent neurological disorders worldwide, characterized by recurrent seizures resulting from abnormal electrical brain activity. Traditional diagnostic methods, such as electroencephalography (EEG) and magnetic resonance imaging (MRI), although widely used, often fail to provide sufficient accuracy when applied in isolation due to their inherent limitations in temporal or spatial resolution. These challenges highlight the need for integrated analytical approaches capable of leveraging complementary modalities for improved diagnostic precision. Objectives: This study aims to address the limitations of single-modality diagnostic techniques by proposing a multimodal deep learning framework that integrates EEG and MRI data to enhance the accuracy of epilepsy detection. The objectives include preprocessing EEG signals to remove artifacts, segmenting MRI scans for structural feature extraction, employing CNNs to model EEG spatio-temporal dependencies, using ViTs to capture hierarchical MRI features, and integrating these representations through a cross-attention fusion mechanism for effective supervised classification. Method: The methodology involves preprocessing EEG signals to eliminate artifacts and ensure clean temporal representation, along with MRI scan segmentation to extract structural information. Convolutional neural networks (CNNs) were used to model spatio-temporal patterns within EEG signals, while vision transformers (ViTs) extracted spatially hierarchical features from MRI data. A cross-attention fusion mechanism was designed to integrate these complementary modalities, and a supervised classifier performed the final prediction. Model training and evaluation were carried out using a high-performance computing setup with GPU acceleration to efficiently handle multimodal data. Results: Experimental results demonstrated that the proposed multimodal deep learning model significantly outperformed state-of-the-art pretrained CNN and ViT baselines across major evaluation metrics, including accuracy, precision, recall, F1-score, and Matthews correlation coefficient. The confusion matrix further confirmed robust differentiation between normal and epileptic classes, emphasizing the reliability of the proposed framework for accurate clinical decision-making. Conclusion: The findings of this study indicate that integrating EEG and MRI data through a multimodal fusion framework provides a substantial improvement in epilepsy detection compared to single-modality approaches. The results highlight the importance of combining complementary temporal and structural information to achieve reliable diagnostic performance. This work contributes to advancing early detection, reducing diagnostic errors, and supporting timely treatment strategies, reinforcing the value of multimodal deep learning approaches in personalized healthcare.
Abstract The timely diagnosis and identification of apple leaf diseases is essential to prevent the spread of diseases and ensure the sound development of the apple industry. Convolutional neural networks (CNNs) have achieved phenomenal success in the area of leaf disease detection, which can greatly benefit the agriculture industry. However, their large size and intricate design continue to pose a challenge when it comes to deploying these models on lightweight devices. Although several successful models (e.g., EfficientNets and MobileNets) have been designed to adapt to resource-constrained devices, these models have not been able to achieve significant results in leaf disease detection tasks and leave a performance gap behind. This research gap has motivated us to develop an apple leaf disease detection model that can not only be deployed on lightweight devices but also outperform existing models. In this work, we propose AppViT, a hybrid vision model, combining the features of convolution blocks and multi-head self-attention, to compete with the best-performing models. Specifically, we begin by introducing the convolution blocks that narrow down the size of the feature maps and help the model encode local features progressively. Then, we stack ViT blocks in combination with convolution blocks, allowing the network to capture non-local dependencies and spatial patterns. Embodied with these designs and a hierarchical structure, AppViT demonstrates excellent performance in apple leaf disease detection tasks. Specifically, it achieves 96.38% precision on Plant Pathology 2021—FGVC8 with about 1.3 million parameters, which is 11.3% and 4.3% more accurate than ResNet-50 and EfficientNet-B3. The precision, recall and F score of our proposed model on Plant Pathology 2021—FGVC8 are 0.967, 0.959, and 0.963 respectively.
apple leaf disease detectionlightweight vision transformerplant disease classificationmulti-head self-attention
The adoption and integration of the Internet of Things (IoT) have become essential for the advancement of many industries, unlocking purposeful connections between objects. However, the surge in IoT adoption and integration has also made it a prime target for malicious attacks. Consequently, ensuring the security of IoT systems and ecosystems has emerged as a crucial research area. Notably, advancements in addressing these security threats include the implementation of intrusion detection systems (IDS), garnering considerable attention within the research community. In this study, and in aim to enhance network anomaly detection, we present a novel intrusion detection approach: the Deep Neural Decision Forest-based IDS (DNDF-IDS). The DNDF-IDS incorporates an improved decision forest model coupled with neural networks to achieve heightened accuracy (ACC). Employing four distinct feature selection methods separately, namely principal component analysis (PCA), LASSO regression (LR), SelectKBest, and Random Forest Feature Importance (RFFI), our objective is to streamline training and prediction processes, enhance overall performance, and identify the most correlated features. Evaluation of our model on three diverse datasets (NSL-KDD, CICIDS2017, and UNSW-NB15) reveals impressive ACC values ranging from 94.09% to 98.84%, depending on the dataset and the feature selection method. Notably, our model achieves a remarkable prediction time of 0.1 ms per record. Comparative analyses with other recent random forest and Convolutional Neural Networks (CNN) based models indicate that our DNDF-IDS performs similarly or even outperforms them in certain instances, particularly when utilizing the top 10 features. One key advantage of our novel model lies in its ability to make accurate predictions with only a few features, showcasing an efficient utilization of computational resources.
intrusion detectionInternet of Things securitynetwork anomaly detectiondecision forest
Expert Systems with Applications · مجلد 255 · 124650-124650
Most existing methods for detecting fraudulent news or other forms of disinformation primarily rely on user profiling or content analysis, determining whether a given article aligns with a user’s stylistic or content-related preferences. This research departs from conventional approaches by concentrating not only on the characteristics of shared content but also on user interactions and the resultant topologies of content propagation trees. In our study, we have introduced a deep learning model based on Graph Convolutional Neural Networks (GCNN) and enhanced with multi-head attention mechanisms. This model leverages a wide range of psycholinguistic attributes extracted from users’ posts, including sentiment, emotional content, linguistic features, personality traits, readability, and communication style. We further enrich these attributes with BERT embeddings to improve textual representation. Our research framework involves creating two distinct graph networks: the user interaction graph and the semantic propagation graph. These graphical representations are essential for visualizing dynamic interactions and patterns of information dissemination among users. For each user, we generate a set of carefully crafted features that capture their unique characteristics and behaviors. These user-specific features are then integrated into the User Interaction Graph, enabling a more nuanced understanding and representation of user interactions within our proposed model. The proposed framework has demonstrated superior performance on benchmark datasets, achieving accuracies and precisions of 91.07% and 91.24% on the FakeNewsNet dataset, and 94.20% and 94.92% on the FibVid dataset, respectively. The results obtained from the experimental evaluation have also revealed the effect of various parameters significant to profile disinformers, signifying a substantial improvement in the field of fake news detection and profiling.