Search Results (2,886)

Tea cultivation faces significant threats from red spider mite (RSM: Oligonychus coffeae) infestations, which reduce yields and economic viability in major tea-producing regions. Current automated detection methods rely on supervised deep learning models requiring extensive labeled data, limiting scalability for smallholder farmers. This article proposes AIM-Net (AI-based Infestation Mapping Network) by evaluating SwAV (Swapping Assignments between Views), a self-supervised learning framework, for classifying RSM infestation severity (Mild, Moderate, Severe) using a geo-referenced, field-acquired dataset of RSM infested tea-leaves, Cam-RSM. The methodology combines SwAV pre-training on unlabeled data with fine-tuning on labeled subsets, employing multi-crop augmentation and online clustering to learn discriminative features without full supervision. Comparative analysis against a fully supervised ResNet-50 baseline utilized 5-fold cross-validation, assessing accuracy, F1-scores, and computational efficiency. Results demonstrate SwAV’s superiority, achieving 98.7% overall accuracy (vs. 92.1% for ResNet-50) and macro-average F1-scores of 98.3% across classes, with a 62% reduction in labeled data requirements. The model showed particular strength in Mild_RSM-class detection (F1-score: 98.5%) and computational efficiency, enabling deployment on edge devices. Statistical validation confirmed significant improvements (p < 0.001) over baseline approaches. These findings establish self-supervised learning as a transformative tool for precision pest management, offering resource-efficient solutions for early infestation detection while maintaining high accuracy. Full article

Objective: To develop and validate a ResNet-50-based deep learning model for automatic detection of osteomyelitis (DFO) in plain radiographs of patients with diabetic foot ulcers (DFUs). Research Design and Methods: This retrospective study included 168 patients with type one or type two diabetes and clinical suspicion of DFO confirmed via a surgical bone biopsy. An experienced clinician and a pretrained ResNet-50 model independently interpreted the radiographs. The model was developed using Python-based frameworks with ChatGPT assistance for coding. The diagnostic performance was assessed against the histopathological findings, calculating sensitivity, specificity, the positive predictive value (PPV), the negative predictive value (NPV), and the likelihood ratios. Agreement between the AI model and the clinician was evaluated using Cohen’s kappa coefficient. Results: The AI model demonstrated high sensitivity (92.8%) and PPV (0.97), but low-level specificity (4.4%). The clinician showed 90.2% sensitivity and 37.8% specificity. The Cohen’s kappa coefficient between the AI model and the clinician was −0.105 (p = 0.117), indicating weak agreement. Both the methods tended to classify many cases as DFO-positive, with 81.5% agreement in the positive cases. Conclusions: This study demonstrates the potential of IA to support the radiographic diagnosis of DFO using a ResNet-50-based deep learning model. AI-assisted radiographic interpretation could enhance early DFO detection, particularly in high-prevalence settings. However, further validation is necessary to improve its specificity and assess its utility in primary care. Full article

In response to the problem of complex and variable loads on wind turbine gearbox bearing in working conditions, as well as the limited amount of sound data making fault identification difficult, this study focuses on sound signals and proposes an intelligent diagnostic method using deep learning. By adding the CBAM module in ResNeXt to enhance the model’s attention to important features and combining it with the Arcloss loss function to make the model learn more discriminative features, the generalization ability of the model is strengthened. We used a fine-tuning transfer learning strategy, transferring pre-trained model parameters to the CBAM-ResNeXt50-ArcLoss model and training with an extracted Mel spectrogram of sound signals to extract and classify audio features of the wind turbine gearbox. Experimental validation of the proposed method on collected sound signals showed its effectiveness and superiority. Compared to CNN, ResNet50, ResNeXt50, and CBAM-ResNet50 methods, the CBAM-ResNeXt50-ArcLoss model achieved improvements of 13.3, 3.6, 2.4, and 1.3, respectively. Through comparison with classical algorithms, we demonstrated that the research method proposed in this study exhibits better diagnostic capability in classifying wind turbine gearbox sound signals. Full article

Specific Emitter Identification (SEI) is a specialized technique for identifying different emitters by analyzing the unique characteristics embedded in received signals, known as Radio Frequency Fingerprints (RFFs), and SEI plays a crucial role in civilian applications. Recently, various SEI methods based on deep learning have been proposed. However, in real-world scenarios, the scarcity of accurately labeled data poses a significant challenge to these methods, which typically rely on large-scale supervised training. To address this issue, we propose a novel SEI framework based on self-supervised contrastive learning. Our approach comprises two stages: an unsupervised pretraining phase that uses contrastive loss to learn discriminative RFF representations from unlabeled data, and a supervised fine-tuning stage regularized through virtual adversarial training (VAT) to improve generalization under limited labels. This framework enables effective feature learning while mitigating overfitting. To validate the effectiveness of the proposed method, we collected real-world satellite navigation signals using a 40-meter antenna and conducted extensive experiments. The results demonstrate that our approach achieves outstanding SEI performance, significantly outperforming several mainstream SEI methods, thereby highlighting the practical potential of contrastive self-supervised learning in satellite transmitter identification. Full article

As Android malware grows increasingly sophisticated, traditional detection methods struggle to keep pace, creating an urgent need for robust, interpretable, and real-time solutions to safeguard mobile ecosystems. This study introduces YoloMal-XAI, a novel deep learning framework that transforms Android application files into RGB image representations by mapping DEX (Dalvik Executable), Manifest.xml, and Resources.arsc files to distinct color channels. Evaluated on the CICMalDroid2020 dataset using YOLO11 pretrained classification models, YoloMal-XAI achieves 99.87% accuracy in binary classification and 99.56% in multi-class classification (Adware, Banking, Riskware, SMS, and Benign). Compared to ResNet-50, GoogLeNet, and MobileNetV2, YOLO11 offers competitive accuracy with at least 7× faster training over 100 epochs. Against YOLOv8, YOLO11 achieves comparable or superior accuracy while reducing training time by up to 3.5×. Cross-corpus validation using Drebin and CICAndMal2017 further confirms the model’s generalization capability on previously unseen malware. An ablation study highlights the value of integrating DEX, Manifest, and Resources components, with the full RGB configuration consistently delivering the best performance. Explainable AI (XAI) techniques—Grad-CAM, Grad-CAM++, Eigen-CAM, and HiRes-CAM—are employed to interpret model decisions, revealing the DEX segment as the most influential component. These results establish YoloMal-XAI as a scalable, efficient, and interpretable framework for Android malware detection, with strong potential for future deployment on resource-constrained mobile devices. Full article

In the era of smart agriculture, accurate identification of plant species is critical for effective crop management, biodiversity monitoring, and the sustainable use of medicinal resources. However, existing deep learning approaches often underperform when applied to fine-grained plant classification tasks due to the lack of domain-specific, high-quality datasets and the limited representational capacity of traditional architectures. This study addresses these challenges by introducing a novel, well-curated leaf image dataset consisting of 39 classes of medicinal and aromatic plants collected from the Aromatic and Medicinal Plant Research Station in Odakkali, Kerala, India. To overcome performance bottlenecks observed with a baseline Convolutional Neural Network (CNN) that achieved only 44.94% accuracy, we progressively enhanced model performance through a series of architectural innovations. These included the use of a pre-trained VGG16 network, data augmentation techniques, and fine-tuning of deeper convolutional layers, followed by the integration of Squeeze-and-Excitation (SE) attention blocks. Ultimately, we propose a hybrid deep learning architecture that combines VGG16 with Batch Normalization, Gated Recurrent Units (GRUs), Transformer modules, and Dilated Convolutions. This final model achieved a peak validation accuracy of 95.24%, significantly outperforming several baseline models, such as custom CNN (44.94%), VGG-19 (59.49%), VGG-16 before augmentation (71.52%), Xception (85.44%), Inception v3 (87.97%), VGG-16 after data augumentation (89.24%), VGG-16 after fine-tuning (90.51%), MobileNetV2 (93.67), and VGG16 with SE block (94.94%). These results demonstrate superior capability in capturing both local textures and global morphological features. The proposed solution not only advances the state of the art in plant classification but also contributes a valuable dataset to the research community. Its real-world applicability spans field-based plant identification, biodiversity conservation, and precision agriculture, offering a scalable tool for automated plant recognition in complex ecological and agricultural environments. Full article

User authentication is crucial for safeguarding access to digital systems and services. Biometric authentication serves as a strong and user-friendly alternative to conventional security methods such as passwords and PINs, which are often susceptible to breaches. This study proposes a deep learning-based multimodal biometric system that combines fingerprint (FP) and finger vein (FV) modalities to improve accuracy and security. The system explores three fusion strategies: feature-level fusion (combining feature vectors from each modality), score-level fusion (integrating prediction scores from each modality), and a hybrid approach that leverages both feature and score information. The implementation involved five pretrained convolutional neural network (CNN) models: two unimodal (FP-only and FV-only) and three multimodal models corresponding to each fusion strategy. The models were assessed using the NUPT-FPV dataset, which consists of 33,600 images collected from 140 subjects with a dual-mode acquisition device in varied environmental conditions. The results indicate that the hybrid-level fusion with a dominant score weight (0.7 score, 0.3 feature) achieved the highest accuracy (99.79%) and the lowest equal error rate (EER = 0.0018), demonstrating superior robustness. Overall, the results demonstrate that integrating deep learning with multimodal fusion is highly effective for advancing scalable and accurate biometric authentication solutions suitable for real-world deployments. Full article

Objectives: Automatic tip localization is critical in ultrasound (US)-guided interventions. Although deep learning (DL) has been widely used for precise tip detection, existing methods are limited by the availability of real puncture data and expert annotations. Methods: To address these challenges, we propose a novel method that uses synthetic US puncture data to pre-train DL-based tip detectors, improving their generalization. Synthetic data are generated by fusing clinical US images of healthy controls with tips created using generative DL models. To ensure clinical diversity, we constructed a dataset from scans of 20 volunteers, covering 20 organs or anatomical regions, obtained with six different US machines and performed by three physicians with varying expertise levels. Tip diversity is introduced by generating a wide range of synthetic tips using a denoising probabilistic diffusion model (DDPM). This method synthesizes a large volume of diverse US puncture data, which are used to pre-train tip detectors, followed by subsequently training with real puncture data. Results: Our method outperforms MSCOCO pre-training on a clinical puncture dataset, achieving a 1.27–7.19% improvement in AP_0.1:0.5 with varying numbers of real samples. State-of-the-art detectors also show performance gains of 1.14–1.76% when applying the proposed method. Conclusions: The experimental results demonstrate that our method enhances the generalization of tip detectors without relying on expert annotations or large amounts of real data, offering significant potential for more accurate visual guidance during US-guided interventions and broader clinical applications. Full article

Background/Objectives: Melanoma is an aggressive type of skin cancer that poses serious health risks if not detected in its early stages. Although early diagnosis enables effective treatment, delays can result in life-threatening consequences. Traditional diagnostic processes predominantly rely on the subjective expertise of dermatologists, which can lead to variability and time inefficiencies. Consequently, there is an increasing demand for automated systems that can accurately classify melanoma lesions and retrieve visually similar cases to support clinical decision-making. Methods: This study proposes a transfer learning (TL)-based deep learning (DL) framework for the classification of melanoma images and the enhancement of content-based image retrieval (CBIR) systems. Pre-trained models including DenseNet121, InceptionV3, Vision Transformer (ViT), and Xception were employed to extract deep feature representations. These features were integrated using a weighted fusion strategy and classified through an Ensemble learning approach designed to capitalize on the complementary strengths of the individual models. The performance of the proposed system was evaluated using classification accuracy and mean Average Precision (mAP) metrics. Results: Experimental evaluations demonstrated that the proposed Ensemble model significantly outperformed each standalone model in both classification and retrieval tasks. The Ensemble approach achieved a classification accuracy of 95.25%. In the CBIR task, the system attained a mean Average Precision (mAP) score of 0.9538, indicating high retrieval effectiveness. The performance gains were attributed to the synergistic integration of features from diverse model architectures through the ensemble and fusion strategies. Conclusions: The findings underscore the effectiveness of TL-based DL models in automating melanoma image classification and enhancing CBIR systems. The integration of deep features from multiple pre-trained models using an Ensemble approach not only improved accuracy but also demonstrated robustness in feature generalization. This approach holds promise for integration into clinical workflows, offering improved diagnostic accuracy and efficiency in the early detection of melanoma. Full article

High-quality, openly accessible clinical datasets remain a significant bottleneck in advancing both research and clinical applications within medical artificial intelligence. Case reports, often rich in multimodal clinical data, represent an underutilized resource for developing medical AI applications. We present an enhanced version of MultiCaRe, a dataset derived from open-access case reports on PubMed Central. This new version addresses the limitations identified in the previous release and incorporates newly added clinical cases and images (totaling 93,816 and 130,791, respectively), along with a refined hierarchical taxonomy featuring over 140 categories. Image labels have been meticulously curated using a combination of manual and machine learning-based label generation and validation, ensuring a higher quality for image classification tasks and the fine-tuning of multimodal models. To facilitate its use, we also provide a Python package for dataset manipulation, pretrained models for medical image classification, and two dedicated websites. The updated MultiCaRe dataset expands the resources available for multimodal AI research in medicine. Its scale, quality, and accessibility make it a valuable tool for developing medical AI systems, as well as for educational purposes in clinical and computational fields. Full article

Antimicrobial peptides (AMPs) provide a robust alternative to conventional antibiotics, combating escalating microbial resistance through their diverse functions and broad pathogen-targeting abilities. While current deep learning technologies enhance AMP generation, they face challenges in developing multifunctional AMPs due to intricate amino acid interdependencies and limited consideration of diverse functional activities. To overcome this challenge, we introduce a novel de novo multifunctional AMP design framework that enhances a Feedback Generative Adversarial Network (FBGAN) by integrating a global quantitative AMP activity regression module and a multifunctional-attribute integrated prediction module. This integrated approach not only facilitates the automated generation of potential AMP candidates, but also optimizes the network’s ability to assess their multifunctionality. Initially, by integrating an effective pre-trained regression and classification model with feedback-loop mechanisms, our model can not only identify potential valid AMP candidates, but also optimizes computational predictions of Minimum Inhibitory Concentration (MIC) values. Subsequently, we employ a combinatorial predictor to simultaneously identify and predict five multifunctional AMP bioactivities, enabling the generation of multifunctional AMPs. The experimental results demonstrate the efficiency of generating AMPs with multiple enhanced antimicrobial properties, indicating that our work can provide a valuable reference for combating multi-drug-resistant infections. Full article

This paper presents the implementation of a language model–based strategy for the automatic codification of crime narratives for the production of official statistics. To address the high workload and inconsistencies associated with manual coding, we developed and evaluated three models: an XGBoost classifier with bag-of-words features and word embeddings features, an LSTM network using pretrained Spanish word embeddings as a language model, and a fine-tuned BERT language model (BETO). Deep learning models outperformed the traditional baseline, with BETO achieving the highest accuracy. The new ENUSC (Encuesta Nacional Urbana de Seguridad Ciudadana) workflow integrates the selected model into an API for automated classification, incorporating a certainty threshold to distinguish between cases suitable for automation and those requiring expert review. This hybrid strategy led to a 68.4% reduction in manual review workload while preserving high-quality standards. This study represents the first documented application of deep learning for the automated classification of victimization narratives in official statistics, demonstrating its feasibility and impact in a real-world production environment. Our results demonstrate that deep learning can significantly improve the efficiency and consistency of crime statistics coding, offering a scalable solution for other national statistical offices. Full article

Purpose: Predicting colorectal cancer liver metastasis (CRLM) is essential for prognostic assessment. This study aims to develop and validate an interpretable multimodal machine learning framework based on multiparametric MRI for predicting CRLM, and to enhance the clinical interpretability of the model through SHapley Additive exPlanations (SHAP) analysis and deep learning visualization. Methods: This multicenter retrospective study included 463 patients with pathologically confirmed colorectal cancer from two institutions, divided into training (n = 256), internal testing (n = 111), and external validation (n = 96) sets. Radiomics features were extracted from manually segmented regions on axial T2-weighted imaging (T2WI) and diffusion-weighted imaging (DWI). Deep learning features were obtained from a pretrained ResNet101 network using the same MRI inputs. A least absolute shrinkage and selection operator (LASSO) logistic regression classifier was developed for clinical, radiomics, deep learning, and combined models. Model performance was evaluated by AUC, sensitivity, specificity, and F1-score. SHAP was used to assess feature contributions, and Grad-CAM was applied to visualize deep feature attention. Results: The combined model integrating features across the three modalities achieved the highest performance across all datasets, with AUCs of 0.889 (training), 0.838 (internal test), and 0.822 (external validation), outperforming single-modality models. Decision curve analysis (DCA) revealed enhanced clinical net benefit from the integrated model, while calibration curves confirmed its good predictive consistency. SHAP analysis revealed that radiomic features related to T2WI texture (e.g., LargeDependenceLowGrayLevelEmphasis) and clinical biomarkers (e.g., CA19-9) were among the most predictive for CRLM. Grad-CAM visualizations confirmed that the deep learning model focused on tumor regions consistent with radiological interpretation. Conclusions: This study presents a robust and interpretable multiparametric MRI-based model for noninvasively predicting liver metastasis in colorectal cancer patients. By integrating handcrafted radiomics and deep learning features, and enhancing transparency through SHAP and Grad-CAM, the model provides both high predictive performance and clinically meaningful explanations. These findings highlight its potential value as a decision-support tool for individualized risk assessment and treatment planning in the management of colorectal cancer. Full article

Machine learning techniques have revolutionized physical layer security (PLS) and provided opportunities for optimizing the performance and security of modern communication systems. In this study, we propose the first machine learning-assisted random communication system (ML-RCS). It comprises a pretrained decision tree (DT)-based receiver that extracts binary information from the transmitted random noise carrier signals. The ML-RCS employs skewed alpha-stable ($\alpha$-stable) noise as a random carrier to encode the incoming binary bits securely. The DT model is pretrained on an extensively developed dataset encompassing all the selected parameter combinations to generate and detect the $\alpha$-stable noise signals. The legitimate receiver leverages the pretrained DT and a predetermined key, specifically the pulse length of a single binary information bit, to securely decode the hidden binary bits. The performance evaluations included the single-bit transmission, confusion matrices, and a bit error rate (BER) analysis via Monte Carlo simulations. The fact that the BER reached 10⁻³ confirms the ability of the proposed system to establish successful secure communication between a transmitter and legitimate receiver. Additionally, the ML-RCS provides an increased data rate compared to previous random communication systems. From the perspective of security, the confusion matrices and computed false negative rate of 50.2% demonstrate the failure of an eavesdropper to decode the binary bits without access to the predetermined key and the private dataset. These findings highlight the potential ability of unconventional ML-RCSs to promote the development of secure next-generation communication devices with built-in PLSs. Full article

Vision-based end-to-end navigation systems have shown impressive capabilities, especially when combined with Imitation Learning (IL) and advanced Deep Learning architectures, such as Transformers. One such example is CIL++, a Transformer-based architecture that learns to map navigation states to vehicle controls based on expert demonstrations only. Nevertheless, reliance on experts’ datasets limits generalization and can lead to failures in unknown circumstances. Deep Reinforcement Learning (DRL) can address this issue by fine-tuning the pretrained policy, using a reward function that aims to improve its weaknesses through interaction with the environment. However, fine-tuning with DRL can lead to the Catastrophic Forgetting (CF) problem, where a policy forgets the expert behaviors learned from the demonstrations as it learns to optimize the new reward function. In this paper, we present CILRLv3, a DRL-based training method that is immune to CF, enabling pretrained navigation agents to improve their driving skills across new scenarios. Full article