In modern wireless communication and electromagnetic control,automatic modulationclassification(AMC)of orthogonal frequency division multiplexing(OFDM)signals plays animportant role.However,under Doppler frequency shi...In modern wireless communication and electromagnetic control,automatic modulationclassification(AMC)of orthogonal frequency division multiplexing(OFDM)signals plays animportant role.However,under Doppler frequency shift and complex multipath channel conditions,extracting discriminative features from high-order modulation signals and ensuring model inter-pretability remain challenging.To address these issues,this paper proposes a Fourier attention net-work(FAttNet),which combines an attention mechanism with a Fourier analysis network(FAN).Specifically,the method directly converts the input signal to the frequency domain using the FAN,thereby obtaining frequency features that reflect the periodic variations in amplitude and phase.Abuilt-in attention mechanism then automatically calculates the weights for each frequency band,focusing on the most discriminative components.This approach improves both classification accu-racy and model interpretability.Experimental validation was conducted via high-order modulationsimulation using an RF testbed.The results show that under three different Doppler frequencyshifts and complex multipath channel conditions,with a signal-to-noise ratio of 10 dB,the classifi-cation accuracy can reach 89.1%,90.4%and 90%,all of which are superior to the current main-stream methods.The proposed approach offers practical value for dynamic spectrum access and sig-nal security detection,and it makes important theoretical contributions to the application of deeplearning in complex electromagnetic signal recognition.展开更多
Background:Accurate classification of normal blood cells is a critical foundation for automated hematological analysis,including the detection of pathological conditions like leukemia.While convolutional neural networ...Background:Accurate classification of normal blood cells is a critical foundation for automated hematological analysis,including the detection of pathological conditions like leukemia.While convolutional neural networks(CNNs)excel in local feature extraction,their ability to capture global contextual relationships in complex cellular morphologies is limited.This study introduces a hybrid CNN-Transformer framework to enhance normal blood cell classification,laying the groundwork for future leukemia diagnostics.Methods:The proposed architecture integrates pre-trained CNNs(ResNet50,EfficientNetB3,InceptionV3,CustomCNN)with Vision Transformer(ViT)layers to combine local and global feature modeling.Four hybrid models were evaluated on the publicly available Blood Cell Images dataset from Kaggle,comprising 17,092 annotated normal blood cell images across eight classes.The models were trained using transfer learning,fine-tuning,and computational optimizations,including cross-model parameter sharing to reduce redundancy by reusing weights across CNN backbones and attention-guided layer pruning to eliminate low-contribution layers based on attention scores,improving efficiency without sacrificing accuracy.Results:The InceptionV3-ViT model achieved a weighted accuracy of 97.66%(accounting for class imbalance by weighting each class’s contribution),a macro F1-score of 0.98,and a ROC-AUC of 0.998.The framework excelled in distinguishing morphologically similar cell types demonstrating robustness and reliable calibration(ECE of 0.019).The framework addresses generalization challenges,including class imbalance and morphological similarities,ensuring robust performance across diverse cell types.Conclusion:The hybrid CNN-Transformer framework significantly improves normal blood cell classification by capturing multi-scale features and long-range dependencies.Its high accuracy,efficiency,and generalization position it as a strong baseline for automated hematological analysis,with potential for extension to leukemia subtype classification through future validation on pathological samples.展开更多
Objective To determine the correlation between traditional Chinese medicine(TCM)inspec-tion of spirit classification and the severity grade of depression based on facial features,offer-ing insights for intelligent int...Objective To determine the correlation between traditional Chinese medicine(TCM)inspec-tion of spirit classification and the severity grade of depression based on facial features,offer-ing insights for intelligent intergrated TCM and western medicine diagnosis of depression.Methods Using the Audio-Visual Emotion Challenge and Workshop(AVEC 2014)public dataset on depression,which conclude 150 interview videos,the samples were classified ac-cording to the TCM inspection of spirit classification:Deshen(得神,presence of spirit),Shaoshen(少神,insufficiency of spirit),and Shenluan(神乱,confusion of spirit).Meanwhile,based on Beck Depression Inventory-II(BDI-II)score for the severity grade of depression,the samples were divided into minimal(0-13,Q1),mild(14-19,Q2),moderate(20-28,Q3),and severe(29-63,Q4).Sixty-eight landmarks were extracted with a ResNet-50 network,and the feature extracion mode was stadardized.Random forest and support vectior machine(SVM)classifiers were used to predict TCM inspection of spirit classification and the severity grade of depression,respectively.A Chi-square test and Apriori association rule mining were then applied to quantify and explore the relationships.Results The analysis revealed a statistically significant and moderately strong association be-tween TCM spirit classification and the severity grade of depression,as confirmed by a Chi-square test(χ^(2)=14.04,P=0.029)with a Cramer’s V effect size of 0.243.Further exploration us-ing association rule mining identified the most compelling rule:“moderate depression(Q3)→Shenluan”.This rule demonstrated a support level of 5%,indicating this specific co-occur-rence was present in 5%of the cohort.Crucially,it achieved a high Confidence of 86%,mean-ing that among patients diagnosed with Q3,86%exhibited the Shenluan pattern according to TCM assessment.The substantial Lift of 2.37 signifies that the observed likelihood of Shenlu-an manifesting in Q3 patients is 2.37 times higher than would be expected by chance if these states were independent-compelling evidence of a highly non-random association.Conse-quently,Shenluan emerges as a distinct and core TCM diagnostic manifestation strongly linked to Q3,forming a clinically significant phenotype within this patient subgroup.展开更多
The network of Himalayan roadways and highways connects some remote regions of valleys or hill slopes,which is vital for India’s socio-economic growth.Due to natural and artificial factors,frequency of slope instabil...The network of Himalayan roadways and highways connects some remote regions of valleys or hill slopes,which is vital for India’s socio-economic growth.Due to natural and artificial factors,frequency of slope instabilities along the networks has been increasing over last few decades.Assessment of stability of natural and artificial slopes due to construction of these connecting road networks is significant in safely executing these roads throughout the year.Several rock mass classification methods are generally used to assess the strength and deformability of rock mass.This study assesses slope stability along the NH-1A of Ramban district of North Western Himalayas.Various structurally and non-structurally controlled rock mass classification systems have been applied to assess the stability conditions of 14 slopes.For evaluating the stability of these slopes,kinematic analysis was performed along with geological strength index(GSI),rock mass rating(RMR),continuous slope mass rating(CoSMR),slope mass rating(SMR),and Q-slope in the present study.The SMR gives three slopes as completely unstable while CoSMR suggests four slopes as completely unstable.The stability of all slopes was also analyzed using a design chart under dynamic and static conditions by slope stability rating(SSR)for the factor of safety(FoS)of 1.2 and 1 respectively.Q-slope with probability of failure(PoF)1%gives two slopes as stable slopes.Stable slope angle has been determined based on the Q-slope safe angle equation and SSR design chart based on the FoS.The value ranges given by different empirical classifications were RMR(37-74),GSI(27.3-58.5),SMR(11-59),and CoSMR(3.39-74.56).Good relationship was found among RMR&SSR and RMR&GSI with correlation coefficient(R 2)value of 0.815 and 0.6866,respectively.Lastly,a comparative stability of all these slopes based on the above classification has been performed to identify the most critical slope along this road.展开更多
Objective:To explore the feasibility of remotely obtaining complex information on traditional Chinese medicine(TCM)pulse conditions through voice signals.Methods: We used multi-label pulse conditions as the entry poin...Objective:To explore the feasibility of remotely obtaining complex information on traditional Chinese medicine(TCM)pulse conditions through voice signals.Methods: We used multi-label pulse conditions as the entry point and modeled and analyzed TCM pulse diagnosis by combining voice analysis and machine learning.Audio features were extracted from voice recordings in the TCM pulse condition dataset.The obtained features were combined with information from tongue and facial diagnoses.A multi-label pulse condition voice classification DNN model was built using 10-fold cross-validation,and the modeling methods were validated using publicly available datasets.Results: The analysis showed that the proposed method achieved an accuracy of 92.59%on the public dataset.The accuracies of the three single-label pulse manifestation models in the test set were 94.27%,96.35%,and 95.39%.The absolute accuracy of the multi-label model was 92.74%.Conclusion: Voice data analysis may serve as a remote adjunct to the TCM diagnostic method for pulse condition assessment.展开更多
Effluent outfalls are an important exit for pollutants discharged from the source flowing into environmental water bodies,as well as an important guarantee for the ecological environment of natural water bodies.In res...Effluent outfalls are an important exit for pollutants discharged from the source flowing into environmental water bodies,as well as an important guarantee for the ecological environment of natural water bodies.In response to main problems of large and diverse effluent outfalls,as well as their monitoring analysis,tracing and regulation in China,classification and regulation countermeasures were proposed based on the characteristics of effluent outfalls.It is suggested that a comprehensive management and control system should be built by improving the management and control system,upgrading monitoring techniques and strengthening social supervision and public education,so as to provide a scientific basis for the supervision and management of effluent outfalls in China and help promote the improvement of water quality and the sustainable development and utilization of water resources.展开更多
Due to the complexity of underground engineering geology,the tunnel boring machine(TBM)usually shows poor adaptability to the surrounding rock mass,leading to machine jamming and geological hazards.For the TBM project...Due to the complexity of underground engineering geology,the tunnel boring machine(TBM)usually shows poor adaptability to the surrounding rock mass,leading to machine jamming and geological hazards.For the TBM project of Lanzhou Water Source Construction,this study proposed a neural network called PCA-GRU,which combines principal component analysis(PCA)with gated recurrent unit(GRU)to improve the accuracy of predicting rock mass classification in TBM tunneling.The input variables from the PCA dimension reduction of nine parameters in the sample data set were utilized for establishing the PCA-GRU model.Subsequently,in order to speed up the response time of surrounding rock mass classification predictions,the PCA-GRU model was optimized.Finally,the prediction results obtained by the PCA-GRU model were compared with those of four other models and further examined using random sampling analysis.As indicated by the results,the PCA-GRU model can predict the rock mass classification in TBM tunneling rapidly,requiring about 20 s to run.It performs better than the previous four models in predicting the rock mass classification,with accuracy A,macro precision MP,and macro recall MR being 0.9667,0.963,and 0.9763,respectively.In Class II,III,and IV rock mass prediction,the PCA-GRU model demonstrates better precision P and recall R owing to the dimension reduction technique.The random sampling analysis indicates that the PCA-GRU model shows stronger generalization,making it more appropriate in situations where the distribution of various rock mass classes and lithologies change in percentage.展开更多
This study investigates the use of a decision tree classification model, combined with Principal Component Analysis (PCA), to distinguish between Assam and Bhutan ethnic groups based on specific anthropometric feature...This study investigates the use of a decision tree classification model, combined with Principal Component Analysis (PCA), to distinguish between Assam and Bhutan ethnic groups based on specific anthropometric features, including age, height, tail length, hair length, bang length, reach, and earlobe type. The dataset was reduced using PCA, which identified height, reach, and age as key features contributing to variance. However, while PCA effectively reduced dimensionality, it faced challenges in clearly distinguishing between the two ethnic groups, a limitation noted in previous research. In contrast, the decision tree model performed significantly better, establishing clear decision boundaries and achieving high classification accuracy. The decision tree consistently selected Height and Reach as the most important classifiers, a finding supported by existing studies on ethnic differences in Northeast India. The results highlight the strengths of combining PCA for dimensionality reduction with decision tree models for classification tasks. While PCA alone was insufficient for optimal class separation, its integration with decision trees improved both the model’s accuracy and interpretability. Future research could explore other machine learning models to enhance classification and examine a broader set of anthropometric features for more comprehensive ethnic group classification.展开更多
The integration of image analysis through deep learning(DL)into rock classification represents a significant leap forward in geological research.While traditional methods remain invaluable for their expertise and hist...The integration of image analysis through deep learning(DL)into rock classification represents a significant leap forward in geological research.While traditional methods remain invaluable for their expertise and historical context,DL offers a powerful complement by enhancing the speed,objectivity,and precision of the classification process.This research explores the significance of image data augmentation techniques in optimizing the performance of convolutional neural networks(CNNs)for geological image analysis,particularly in the classification of igneous,metamorphic,and sedimentary rock types from rock thin section(RTS)images.This study primarily focuses on classic image augmentation techniques and evaluates their impact on model accuracy and precision.Results demonstrate that augmentation techniques like Equalize significantly enhance the model's classification capabilities,achieving an F1-Score of 0.9869 for igneous rocks,0.9884 for metamorphic rocks,and 0.9929 for sedimentary rocks,representing improvements compared to the baseline original results.Moreover,the weighted average F1-Score across all classes and techniques is 0.9886,indicating an enhancement.Conversely,methods like Distort lead to decreased accuracy and F1-Score,with an F1-Score of 0.949 for igneous rocks,0.954 for metamorphic rocks,and 0.9416 for sedimentary rocks,exacerbating the performance compared to the baseline.The study underscores the practicality of image data augmentation in geological image classification and advocates for the adoption of DL methods in this domain for automation and improved results.The findings of this study can benefit various fields,including remote sensing,mineral exploration,and environmental monitoring,by enhancing the accuracy of geological image analysis both for scientific research and industrial applications.展开更多
Machine learning(ML)is increasingly applied for medical image processing with appropriate learning paradigms.These applications include analyzing images of various organs,such as the brain,lung,eye,etc.,to identify sp...Machine learning(ML)is increasingly applied for medical image processing with appropriate learning paradigms.These applications include analyzing images of various organs,such as the brain,lung,eye,etc.,to identify specific flaws/diseases for diagnosis.The primary concern of ML applications is the precise selection of flexible image features for pattern detection and region classification.Most of the extracted image features are irrelevant and lead to an increase in computation time.Therefore,this article uses an analytical learning paradigm to design a Congruent Feature Selection Method to select the most relevant image features.This process trains the learning paradigm using similarity and correlation-based features over different textural intensities and pixel distributions.The similarity between the pixels over the various distribution patterns with high indexes is recommended for disease diagnosis.Later,the correlation based on intensity and distribution is analyzed to improve the feature selection congruency.Therefore,the more congruent pixels are sorted in the descending order of the selection,which identifies better regions than the distribution.Now,the learning paradigm is trained using intensity and region-based similarity to maximize the chances of selection.Therefore,the probability of feature selection,regardless of the textures and medical image patterns,is improved.This process enhances the performance of ML applications for different medical image processing.The proposed method improves the accuracy,precision,and training rate by 13.19%,10.69%,and 11.06%,respectively,compared to other models for the selected dataset.The mean error and selection time is also reduced by 12.56%and 13.56%,respectively,compared to the same models and dataset.展开更多
Autonomous Underwater Vehicles(AUVs)are pivotal for deep-sea exploration and resource exploitation,yet their reliability in extreme underwater environments remains a critical barrier to widespread deployment.Through s...Autonomous Underwater Vehicles(AUVs)are pivotal for deep-sea exploration and resource exploitation,yet their reliability in extreme underwater environments remains a critical barrier to widespread deployment.Through systematic analysis of 150 peer-reviewed studies employing mixed-methods research,this review yields three principal advancements to the reliability analysis of AUVs.First,based on the hierarchical functional division of AUVs into six subsystems(propulsion system,navigation system,communication system,power system,environmental detection system,and emergency system),this study systematically identifies the primary failure modes and potential failure causes of each subsystem,providing theoretical support for fault diagnosis and reliability optimization.Subsequently,a comprehensive review of AUV reliability analysis methods is conducted from three perspectives:analytical methods,simulated methods,and surrogate model methods.The applicability and limitations of each method are critically analyzed to offer insights into their suitability for engineering applications.Finally,the study highlights key challenges and research hotpots in AUV reliability analysis,including reliability analysis under limited data,AI-driven reliability analysis,and human reliability analysis.Furthermore,the potential of multi-sensor data fusion,edge computing,and advanced materials in enhancing AUV environmental adaptability and reliability is explored.展开更多
Cyberbullying on social media poses significant psychological risks,yet most detection systems over-simplify the task by focusing on binary classification,ignoring nuanced categories like passive-aggressive remarks or...Cyberbullying on social media poses significant psychological risks,yet most detection systems over-simplify the task by focusing on binary classification,ignoring nuanced categories like passive-aggressive remarks or indirect slurs.To address this gap,we propose a hybrid framework combining Term Frequency-Inverse Document Frequency(TF-IDF),word-to-vector(Word2Vec),and Bidirectional Encoder Representations from Transformers(BERT)based models for multi-class cyberbullying detection.Our approach integrates TF-IDF for lexical specificity and Word2Vec for semantic relationships,fused with BERT’s contextual embeddings to capture syntactic and semantic complexities.We evaluate the framework on a publicly available dataset of 47,000 annotated social media posts across five cyberbullying categories:age,ethnicity,gender,religion,and indirect aggression.Among BERT variants tested,BERT Base Un-Cased achieved the highest performance with 93%accuracy(standard deviation across±1%5-fold cross-validation)and an average AUC of 0.96,outperforming standalone TF-IDF(78%)and Word2Vec(82%)models.Notably,it achieved near-perfect AUC scores(0.99)for age and ethnicity-based bullying.A comparative analysis with state-of-the-art benchmarks,including Generative Pre-trained Transformer 2(GPT-2)and Text-to-Text Transfer Transformer(T5)models highlights BERT’s superiority in handling ambiguous language.This work advances cyberbullying detection by demonstrating how hybrid feature extraction and transformer models improve multi-class classification,offering a scalable solution for moderating nuanced harmful content.展开更多
In the rapidly evolving landscape of natural language processing(NLP)and sentiment analysis,improving the accuracy and efficiency of sentiment classification models is crucial.This paper investigates the performance o...In the rapidly evolving landscape of natural language processing(NLP)and sentiment analysis,improving the accuracy and efficiency of sentiment classification models is crucial.This paper investigates the performance of two advanced models,the Large Language Model(LLM)LLaMA model and NLP BERT model,in the context of airline review sentiment analysis.Through fine-tuning,domain adaptation,and the application of few-shot learning,the study addresses the subtleties of sentiment expressions in airline-related text data.Employing predictive modeling and comparative analysis,the research evaluates the effectiveness of Large Language Model Meta AI(LLaMA)and Bidirectional Encoder Representations from Transformers(BERT)in capturing sentiment intricacies.Fine-tuning,including domain adaptation,enhances the models'performance in sentiment classification tasks.Additionally,the study explores the potential of few-shot learning to improve model generalization using minimal annotated data for targeted sentiment analysis.By conducting experiments on a diverse airline review dataset,the research quantifies the impact of fine-tuning,domain adaptation,and few-shot learning on model performance,providing valuable insights for industries aiming to predict recommendations and enhance customer satisfaction through a deeper understanding of sentiment in user-generated content(UGC).This research contributes to refining sentiment analysis models,ultimately fostering improved customer satisfaction in the airline industry.展开更多
Traffic encryption techniques facilitate cyberattackers to hide their presence and activities.Traffic classification is an important method to prevent network threats.However,due to the tremendous traffic volume and l...Traffic encryption techniques facilitate cyberattackers to hide their presence and activities.Traffic classification is an important method to prevent network threats.However,due to the tremendous traffic volume and limitations of computing,most existing traffic classification techniques are inapplicable to the high-speed network environment.In this paper,we propose a High-speed Encrypted Traffic Classification(HETC)method containing two stages.First,to efficiently detect whether traffic is encrypted,HETC focuses on randomly sampled short flows and extracts aggregation entropies with chi-square test features to measure the different patterns of the byte composition and distribution between encrypted and unencrypted flows.Second,HETC introduces binary features upon the previous features and performs fine-grained traffic classification by combining these payload features with a Random Forest model.The experimental results show that HETC can achieve a 94%F-measure in detecting encrypted flows and a 85%–93%F-measure in classifying fine-grained flows for a 1-KB flow-length dataset,outperforming the state-of-the-art comparison methods.Meanwhile,HETC does not need to wait for the end of the flow and can extract mass computing features.The average time for HETC to process each flow is only 2 or 16 ms,which is lower than the flow duration in most cases,making it a good candidate for high-speed traffic classification.展开更多
The early and precise identification of Alzheimer’s Disease(AD)continues to pose considerable clinical difficulty due to subtle structural alterations and overlapping symptoms across the disease phases.This study pre...The early and precise identification of Alzheimer’s Disease(AD)continues to pose considerable clinical difficulty due to subtle structural alterations and overlapping symptoms across the disease phases.This study presents a novel Deformable Attention Vision Transformer(DA-ViT)architecture that integrates deformable Multi-Head Self-Attention(MHSA)with a Multi-Layer Perceptron(MLP)block for efficient classification of Alzheimer’s disease(AD)using Magnetic resonance imaging(MRI)scans.In contrast to traditional vision transformers,our deformable MHSA module preferentially concentrates on spatially pertinent patches through learned offset predictions,markedly diminishing processing demands while improving localized feature representation.DA-ViT contains only 0.93 million parameters,making it exceptionally suitable for implementation in resource-limited settings.We evaluate the model using a class-imbalanced Alzheimer’s MRI dataset comprising 6400 images across four categories,achieving a test accuracy of 80.31%,a macro F1-score of 0.80,and an area under the receiver operating characteristic curve(AUC)of 1.00 for the Mild Demented category.Thorough ablation studies validate the ideal configuration of transformer depth,headcount,and embedding dimensions.Moreover,comparison research indicates that DA-ViT surpasses state-of-theart pre-trained Convolutional Neural Network(CNN)models in terms of accuracy and parameter efficiency.展开更多
Background:Early and accurate diagnosis of cataracts,which ranks among the leading preventable causes of blindness,is critical to securing positive outcomes for patients.Recently,eye image analyses have used deep lear...Background:Early and accurate diagnosis of cataracts,which ranks among the leading preventable causes of blindness,is critical to securing positive outcomes for patients.Recently,eye image analyses have used deep learning(DL)approaches to automate cataract classification more precisely,leading to the development of the Multiscale Parallel Feature Aggregation Network with Attention Fusion(MPFAN-AF).Focused on improving a model’s performance,this approach applies multiscale feature extraction,parallel feature fusion,along with attention-based fusion to sharpen its focus on salient features,which are crucial in detecting cataracts.Methods:Coarse-level features are captured through the application of convolutional layers,and these features undergo refinement through layered kernels of varying sizes.Moreover,this method captures all the diverse representations of cataracts accurately by parallel feature aggregation.Utilizing the Cataract Eye Dataset available on Kaggle,containing 612 labelled images of eyes with and without cataracts proportionately(normal vs.pathological),this model was trained and tested.Results:Results using the proposed model reflect greater precision over traditional convolutional neural networks(CNNs)models,achieving a classification accuracy of 97.52%.Additionally,the model demonstrated exceptional performance in classification tasks.The ablation studies validated that all applications added value to the prediction process,particularly emphasizing the attention fusion module.Conclusion:The MPFAN-AF model demonstrates high efficiency together with interpretability because it shows promise as an integration solution for real-time mobile cataract detection screening systems.Standard performance indicators indicate that AI-based ophthalmology tools have a promising future for use in remote conditions that lack medical resources.展开更多
The endpoint carbon content in the converter is critical for the quality of steel products,and accurately predicting this parameter is an effective way to reduce alloy consumption and improve smelting efficiency.Howev...The endpoint carbon content in the converter is critical for the quality of steel products,and accurately predicting this parameter is an effective way to reduce alloy consumption and improve smelting efficiency.However,most scholars currently focus on modifying methods to enhance model accuracy,while overlooking the extent to which input parameters influence accuracy.To address this issue,in this study,a prediction model for the endpoint carbon content in the converter was developed using factor analysis(FA)and support vector machine(SVM)optimized by improved particle swarm optimization(IPSO).Analysis of the factors influencing the endpoint carbon content during the converter smelting process led to the identification of 21 input parameters.Subsequently,FA was used to reduce the dimensionality of the data and applied to the prediction model.The results demonstrate that the performance of the FA-IPSO-SVM model surpasses several existing methods,such as twin support vector regression and support vector machine.The model achieves hit rates of 89.59%,96.21%,and 98.74%within error ranges of±0.01%,±0.015%,and±0.02%,respectively.Finally,based on the prediction results obtained by sequentially removing input parameters,the parameters were classified into high influence(5%-7%),medium influence(2%-5%),and low influence(0-2%)categories according to their varying degrees of impact on prediction accuracy.This classi-fication provides a reference for selecting input parameters in future prediction models for endpoint carbon content.展开更多
Unmanned Aerial Vehicles(UAVs)have become indispensable for intelligent traffic monitoring,particularly in low-light conditions,where traditional surveillance systems struggle.This study presents a novel deep learning...Unmanned Aerial Vehicles(UAVs)have become indispensable for intelligent traffic monitoring,particularly in low-light conditions,where traditional surveillance systems struggle.This study presents a novel deep learning-based framework for nighttime aerial vehicle detection and classification that addresses critical challenges of poor illumination,noise,and occlusions.Our pipeline integrates MSRCR enhancement with OPTICS segmentation to overcome low-light challenges,while YOLOv10 enables accurate vehicle localization.The framework employs GLOH and Dense-SIFT for discriminative feature extraction,optimized using the Whale Optimization Algorithm to enhance classification performance.A Swin Transformer-based classifier provides the final categorization,leveraging hierarchical attention mechanisms for robust performance.Extensive experimentation validates our approach,achieving detection mAP@0.5 scores of 91.5%(UAVDT)and 89.7%(VisDrone),alongside classification accuracies of 95.50%and 92.67%,respectively.These results outperform state-of-the-art methods by up to 5.10%in accuracy and 4.2%in mAP,demonstrating the framework’s effectiveness for real-time aerial surveillance and intelligent traffic management in challenging nighttime environments.展开更多
If progress is to be made toward improving geohazard management and emergency decision-making,then lessons need to be learned from past geohazard information.A geologic hazard report provides a useful and reliable sou...If progress is to be made toward improving geohazard management and emergency decision-making,then lessons need to be learned from past geohazard information.A geologic hazard report provides a useful and reliable source of information about the occurrence of an event,along with detailed information about the condition or factors of the geohazard.Analyzing such reports,however,can be a challenging process because these texts are often presented in unstructured long text formats,and contain rich specialized and detailed information.Automatically text classification is commonly used to mine disaster text data in open domains(e.g.,news and microblogs).But it has limitations to performing contextual long-distance dependencies and is insensitive to discourse order.These deficiencies are most obviously exposed in long text fields.Therefore,this paper uses the bidirectional encoder representations from Transformers(BERT),to model long text.Then,utilizing a softmax layer to automatically extract text features and classify geohazards without manual features.The latent Dirichlet allocation(LDA)model is used to examine the interdependencies that exist between causal variables to visualize geohazards.The proposed method is useful in enabling the machine-assisted interpretation of text-based geohazards.Moreover,it can help users visualize causes,processes,and other geohazards and assist decision-makers in emergency responses.展开更多
Charcot-Marie-Tooth （CMT） disease is the most common inherited peripheral neuropathic disorder. CMT is clinically and genetically heterogeneous. To date, 27 genes associated with the disease have been cloned. The pr...Charcot-Marie-Tooth （CMT） disease is the most common inherited peripheral neuropathic disorder. CMT is clinically and genetically heterogeneous. To date, 27 genes associated with the disease have been cloned. The present study carried out clinical classification according to clinical, electrophysiological and pathological features, conducted inheritance classification according to inheritance patterns, and performed mutation analysis of 13 CMT disease genes （PMP22, CX32, HSPB1, MNF2, MPZ, HSPB8, GDAP1, NFL, EGR2, SIMPLE, RAB7, LMNA, MTMR2） in 57 Chinese probands with CMT. Five cases of AD-CMT1 and 13 cases of sporadic CMT1 were diagnosed as CMT1A; five cases of X-CMT1, one case of X-CMT2 and one case of sporadic CMT1 were diagnosed as CMTXl; four cases of AD-CMT2 were diagnosed as CMT2F; one case of AD-CMT2 and one case of sporadic CMT2 were diagnosed as CMT2A2; one case of AD-CMT2 was diagnosed as CMT2L; one case of AD-CMT2 was diagnosed as CMT2J; one case of AR-CMT1 was diagnosed as CMT4A. Among the 57 CMT probands, seven genotypes were determined among 34 patients, with a detection rate of 59.6%. The results indicated that the clinical classification and inheritance classification are indispensable for selecting potential disease genes for mutation detection, and for efficient molecular diagnosis.展开更多
基金supported by the National Natural Science Foundation of China(No.62027801).
文摘In modern wireless communication and electromagnetic control,automatic modulationclassification(AMC)of orthogonal frequency division multiplexing(OFDM)signals plays animportant role.However,under Doppler frequency shift and complex multipath channel conditions,extracting discriminative features from high-order modulation signals and ensuring model inter-pretability remain challenging.To address these issues,this paper proposes a Fourier attention net-work(FAttNet),which combines an attention mechanism with a Fourier analysis network(FAN).Specifically,the method directly converts the input signal to the frequency domain using the FAN,thereby obtaining frequency features that reflect the periodic variations in amplitude and phase.Abuilt-in attention mechanism then automatically calculates the weights for each frequency band,focusing on the most discriminative components.This approach improves both classification accu-racy and model interpretability.Experimental validation was conducted via high-order modulationsimulation using an RF testbed.The results show that under three different Doppler frequencyshifts and complex multipath channel conditions,with a signal-to-noise ratio of 10 dB,the classifi-cation accuracy can reach 89.1%,90.4%and 90%,all of which are superior to the current main-stream methods.The proposed approach offers practical value for dynamic spectrum access and sig-nal security detection,and it makes important theoretical contributions to the application of deeplearning in complex electromagnetic signal recognition.
基金the Deanship of Graduate Studies and Scientific Research at Najran University,Saudi Arabia,for their financial support through the Easy Track Research program,grant code(NU/EFP/MRC/13).
文摘Background:Accurate classification of normal blood cells is a critical foundation for automated hematological analysis,including the detection of pathological conditions like leukemia.While convolutional neural networks(CNNs)excel in local feature extraction,their ability to capture global contextual relationships in complex cellular morphologies is limited.This study introduces a hybrid CNN-Transformer framework to enhance normal blood cell classification,laying the groundwork for future leukemia diagnostics.Methods:The proposed architecture integrates pre-trained CNNs(ResNet50,EfficientNetB3,InceptionV3,CustomCNN)with Vision Transformer(ViT)layers to combine local and global feature modeling.Four hybrid models were evaluated on the publicly available Blood Cell Images dataset from Kaggle,comprising 17,092 annotated normal blood cell images across eight classes.The models were trained using transfer learning,fine-tuning,and computational optimizations,including cross-model parameter sharing to reduce redundancy by reusing weights across CNN backbones and attention-guided layer pruning to eliminate low-contribution layers based on attention scores,improving efficiency without sacrificing accuracy.Results:The InceptionV3-ViT model achieved a weighted accuracy of 97.66%(accounting for class imbalance by weighting each class’s contribution),a macro F1-score of 0.98,and a ROC-AUC of 0.998.The framework excelled in distinguishing morphologically similar cell types demonstrating robustness and reliable calibration(ECE of 0.019).The framework addresses generalization challenges,including class imbalance and morphological similarities,ensuring robust performance across diverse cell types.Conclusion:The hybrid CNN-Transformer framework significantly improves normal blood cell classification by capturing multi-scale features and long-range dependencies.Its high accuracy,efficiency,and generalization position it as a strong baseline for automated hematological analysis,with potential for extension to leukemia subtype classification through future validation on pathological samples.
基金Research and Development Plan of Key Areas of Hunan Science and Technology Department (2022SK2044)Clinical Research Center for Depressive Disorder in Hunan Province (2021SK4022)。
文摘Objective To determine the correlation between traditional Chinese medicine(TCM)inspec-tion of spirit classification and the severity grade of depression based on facial features,offer-ing insights for intelligent intergrated TCM and western medicine diagnosis of depression.Methods Using the Audio-Visual Emotion Challenge and Workshop(AVEC 2014)public dataset on depression,which conclude 150 interview videos,the samples were classified ac-cording to the TCM inspection of spirit classification:Deshen(得神,presence of spirit),Shaoshen(少神,insufficiency of spirit),and Shenluan(神乱,confusion of spirit).Meanwhile,based on Beck Depression Inventory-II(BDI-II)score for the severity grade of depression,the samples were divided into minimal(0-13,Q1),mild(14-19,Q2),moderate(20-28,Q3),and severe(29-63,Q4).Sixty-eight landmarks were extracted with a ResNet-50 network,and the feature extracion mode was stadardized.Random forest and support vectior machine(SVM)classifiers were used to predict TCM inspection of spirit classification and the severity grade of depression,respectively.A Chi-square test and Apriori association rule mining were then applied to quantify and explore the relationships.Results The analysis revealed a statistically significant and moderately strong association be-tween TCM spirit classification and the severity grade of depression,as confirmed by a Chi-square test(χ^(2)=14.04,P=0.029)with a Cramer’s V effect size of 0.243.Further exploration us-ing association rule mining identified the most compelling rule:“moderate depression(Q3)→Shenluan”.This rule demonstrated a support level of 5%,indicating this specific co-occur-rence was present in 5%of the cohort.Crucially,it achieved a high Confidence of 86%,mean-ing that among patients diagnosed with Q3,86%exhibited the Shenluan pattern according to TCM assessment.The substantial Lift of 2.37 signifies that the observed likelihood of Shenlu-an manifesting in Q3 patients is 2.37 times higher than would be expected by chance if these states were independent-compelling evidence of a highly non-random association.Conse-quently,Shenluan emerges as a distinct and core TCM diagnostic manifestation strongly linked to Q3,forming a clinically significant phenotype within this patient subgroup.
文摘The network of Himalayan roadways and highways connects some remote regions of valleys or hill slopes,which is vital for India’s socio-economic growth.Due to natural and artificial factors,frequency of slope instabilities along the networks has been increasing over last few decades.Assessment of stability of natural and artificial slopes due to construction of these connecting road networks is significant in safely executing these roads throughout the year.Several rock mass classification methods are generally used to assess the strength and deformability of rock mass.This study assesses slope stability along the NH-1A of Ramban district of North Western Himalayas.Various structurally and non-structurally controlled rock mass classification systems have been applied to assess the stability conditions of 14 slopes.For evaluating the stability of these slopes,kinematic analysis was performed along with geological strength index(GSI),rock mass rating(RMR),continuous slope mass rating(CoSMR),slope mass rating(SMR),and Q-slope in the present study.The SMR gives three slopes as completely unstable while CoSMR suggests four slopes as completely unstable.The stability of all slopes was also analyzed using a design chart under dynamic and static conditions by slope stability rating(SSR)for the factor of safety(FoS)of 1.2 and 1 respectively.Q-slope with probability of failure(PoF)1%gives two slopes as stable slopes.Stable slope angle has been determined based on the Q-slope safe angle equation and SSR design chart based on the FoS.The value ranges given by different empirical classifications were RMR(37-74),GSI(27.3-58.5),SMR(11-59),and CoSMR(3.39-74.56).Good relationship was found among RMR&SSR and RMR&GSI with correlation coefficient(R 2)value of 0.815 and 0.6866,respectively.Lastly,a comparative stability of all these slopes based on the above classification has been performed to identify the most critical slope along this road.
基金supported by Fundamental Research Funds from the Beijing University of Chinese Medicine(2023-JYB-KYPT-13)the Developmental Fund of Beijing University of Chinese Medicine(2020-ZXFZJJ-083).
文摘Objective:To explore the feasibility of remotely obtaining complex information on traditional Chinese medicine(TCM)pulse conditions through voice signals.Methods: We used multi-label pulse conditions as the entry point and modeled and analyzed TCM pulse diagnosis by combining voice analysis and machine learning.Audio features were extracted from voice recordings in the TCM pulse condition dataset.The obtained features were combined with information from tongue and facial diagnoses.A multi-label pulse condition voice classification DNN model was built using 10-fold cross-validation,and the modeling methods were validated using publicly available datasets.Results: The analysis showed that the proposed method achieved an accuracy of 92.59%on the public dataset.The accuracies of the three single-label pulse manifestation models in the test set were 94.27%,96.35%,and 95.39%.The absolute accuracy of the multi-label model was 92.74%.Conclusion: Voice data analysis may serve as a remote adjunct to the TCM diagnostic method for pulse condition assessment.
文摘Effluent outfalls are an important exit for pollutants discharged from the source flowing into environmental water bodies,as well as an important guarantee for the ecological environment of natural water bodies.In response to main problems of large and diverse effluent outfalls,as well as their monitoring analysis,tracing and regulation in China,classification and regulation countermeasures were proposed based on the characteristics of effluent outfalls.It is suggested that a comprehensive management and control system should be built by improving the management and control system,upgrading monitoring techniques and strengthening social supervision and public education,so as to provide a scientific basis for the supervision and management of effluent outfalls in China and help promote the improvement of water quality and the sustainable development and utilization of water resources.
基金State Key Laboratory of Hydroscience and Hydraulic Engineering of Tsinghua University,Grant/Award Number:2019-KY-03Key Technology of Intelligent Construction of Urban Underground Space of North China University of Technology,Grant/Award Number:110051360022XN108-19+3 种基金Research Start-up Fund Project of North China University of Technology,Grant/Award Number:110051360002Yujie Project of North China University of Technology,Grant/Award Number:216051360020XN199/006National Natural Science Foundation of China,Grant/Award Numbers:51522903,51774184National Key R&D Program of China,Grant/Award Numbers:2018YFC1504801,2018YFC1504902。
文摘Due to the complexity of underground engineering geology,the tunnel boring machine(TBM)usually shows poor adaptability to the surrounding rock mass,leading to machine jamming and geological hazards.For the TBM project of Lanzhou Water Source Construction,this study proposed a neural network called PCA-GRU,which combines principal component analysis(PCA)with gated recurrent unit(GRU)to improve the accuracy of predicting rock mass classification in TBM tunneling.The input variables from the PCA dimension reduction of nine parameters in the sample data set were utilized for establishing the PCA-GRU model.Subsequently,in order to speed up the response time of surrounding rock mass classification predictions,the PCA-GRU model was optimized.Finally,the prediction results obtained by the PCA-GRU model were compared with those of four other models and further examined using random sampling analysis.As indicated by the results,the PCA-GRU model can predict the rock mass classification in TBM tunneling rapidly,requiring about 20 s to run.It performs better than the previous four models in predicting the rock mass classification,with accuracy A,macro precision MP,and macro recall MR being 0.9667,0.963,and 0.9763,respectively.In Class II,III,and IV rock mass prediction,the PCA-GRU model demonstrates better precision P and recall R owing to the dimension reduction technique.The random sampling analysis indicates that the PCA-GRU model shows stronger generalization,making it more appropriate in situations where the distribution of various rock mass classes and lithologies change in percentage.
文摘This study investigates the use of a decision tree classification model, combined with Principal Component Analysis (PCA), to distinguish between Assam and Bhutan ethnic groups based on specific anthropometric features, including age, height, tail length, hair length, bang length, reach, and earlobe type. The dataset was reduced using PCA, which identified height, reach, and age as key features contributing to variance. However, while PCA effectively reduced dimensionality, it faced challenges in clearly distinguishing between the two ethnic groups, a limitation noted in previous research. In contrast, the decision tree model performed significantly better, establishing clear decision boundaries and achieving high classification accuracy. The decision tree consistently selected Height and Reach as the most important classifiers, a finding supported by existing studies on ethnic differences in Northeast India. The results highlight the strengths of combining PCA for dimensionality reduction with decision tree models for classification tasks. While PCA alone was insufficient for optimal class separation, its integration with decision trees improved both the model’s accuracy and interpretability. Future research could explore other machine learning models to enhance classification and examine a broader set of anthropometric features for more comprehensive ethnic group classification.
文摘The integration of image analysis through deep learning(DL)into rock classification represents a significant leap forward in geological research.While traditional methods remain invaluable for their expertise and historical context,DL offers a powerful complement by enhancing the speed,objectivity,and precision of the classification process.This research explores the significance of image data augmentation techniques in optimizing the performance of convolutional neural networks(CNNs)for geological image analysis,particularly in the classification of igneous,metamorphic,and sedimentary rock types from rock thin section(RTS)images.This study primarily focuses on classic image augmentation techniques and evaluates their impact on model accuracy and precision.Results demonstrate that augmentation techniques like Equalize significantly enhance the model's classification capabilities,achieving an F1-Score of 0.9869 for igneous rocks,0.9884 for metamorphic rocks,and 0.9929 for sedimentary rocks,representing improvements compared to the baseline original results.Moreover,the weighted average F1-Score across all classes and techniques is 0.9886,indicating an enhancement.Conversely,methods like Distort lead to decreased accuracy and F1-Score,with an F1-Score of 0.949 for igneous rocks,0.954 for metamorphic rocks,and 0.9416 for sedimentary rocks,exacerbating the performance compared to the baseline.The study underscores the practicality of image data augmentation in geological image classification and advocates for the adoption of DL methods in this domain for automation and improved results.The findings of this study can benefit various fields,including remote sensing,mineral exploration,and environmental monitoring,by enhancing the accuracy of geological image analysis both for scientific research and industrial applications.
基金the Deanship of Scientifc Research at King Khalid University for funding this work through large group Research Project under grant number RGP2/421/45supported via funding from Prince Sattam bin Abdulaziz University project number(PSAU/2024/R/1446)+1 种基金supported by theResearchers Supporting Project Number(UM-DSR-IG-2023-07)Almaarefa University,Riyadh,Saudi Arabia.supported by the Basic Science Research Program through the National Research Foundation of Korea(NRF)funded by the Ministry of Education(No.2021R1F1A1055408).
文摘Machine learning(ML)is increasingly applied for medical image processing with appropriate learning paradigms.These applications include analyzing images of various organs,such as the brain,lung,eye,etc.,to identify specific flaws/diseases for diagnosis.The primary concern of ML applications is the precise selection of flexible image features for pattern detection and region classification.Most of the extracted image features are irrelevant and lead to an increase in computation time.Therefore,this article uses an analytical learning paradigm to design a Congruent Feature Selection Method to select the most relevant image features.This process trains the learning paradigm using similarity and correlation-based features over different textural intensities and pixel distributions.The similarity between the pixels over the various distribution patterns with high indexes is recommended for disease diagnosis.Later,the correlation based on intensity and distribution is analyzed to improve the feature selection congruency.Therefore,the more congruent pixels are sorted in the descending order of the selection,which identifies better regions than the distribution.Now,the learning paradigm is trained using intensity and region-based similarity to maximize the chances of selection.Therefore,the probability of feature selection,regardless of the textures and medical image patterns,is improved.This process enhances the performance of ML applications for different medical image processing.The proposed method improves the accuracy,precision,and training rate by 13.19%,10.69%,and 11.06%,respectively,compared to other models for the selected dataset.The mean error and selection time is also reduced by 12.56%and 13.56%,respectively,compared to the same models and dataset.
基金The National Key R&D Program Projects(Grant No.2022YFC2803601)the Natural Science Foundation of Shandong Province(Grant No.ZR2021YQ29)+1 种基金the Natural Science Foundation of Heilongjiang Province(Grant No.YQ2024E036)the Taishan Scholars Project(Grant No.tsqn202312317).
文摘Autonomous Underwater Vehicles(AUVs)are pivotal for deep-sea exploration and resource exploitation,yet their reliability in extreme underwater environments remains a critical barrier to widespread deployment.Through systematic analysis of 150 peer-reviewed studies employing mixed-methods research,this review yields three principal advancements to the reliability analysis of AUVs.First,based on the hierarchical functional division of AUVs into six subsystems(propulsion system,navigation system,communication system,power system,environmental detection system,and emergency system),this study systematically identifies the primary failure modes and potential failure causes of each subsystem,providing theoretical support for fault diagnosis and reliability optimization.Subsequently,a comprehensive review of AUV reliability analysis methods is conducted from three perspectives:analytical methods,simulated methods,and surrogate model methods.The applicability and limitations of each method are critically analyzed to offer insights into their suitability for engineering applications.Finally,the study highlights key challenges and research hotpots in AUV reliability analysis,including reliability analysis under limited data,AI-driven reliability analysis,and human reliability analysis.Furthermore,the potential of multi-sensor data fusion,edge computing,and advanced materials in enhancing AUV environmental adaptability and reliability is explored.
基金funded by Scientific Research Deanship at University of Hail-Saudi Arabia through Project Number RG-23092.
文摘Cyberbullying on social media poses significant psychological risks,yet most detection systems over-simplify the task by focusing on binary classification,ignoring nuanced categories like passive-aggressive remarks or indirect slurs.To address this gap,we propose a hybrid framework combining Term Frequency-Inverse Document Frequency(TF-IDF),word-to-vector(Word2Vec),and Bidirectional Encoder Representations from Transformers(BERT)based models for multi-class cyberbullying detection.Our approach integrates TF-IDF for lexical specificity and Word2Vec for semantic relationships,fused with BERT’s contextual embeddings to capture syntactic and semantic complexities.We evaluate the framework on a publicly available dataset of 47,000 annotated social media posts across five cyberbullying categories:age,ethnicity,gender,religion,and indirect aggression.Among BERT variants tested,BERT Base Un-Cased achieved the highest performance with 93%accuracy(standard deviation across±1%5-fold cross-validation)and an average AUC of 0.96,outperforming standalone TF-IDF(78%)and Word2Vec(82%)models.Notably,it achieved near-perfect AUC scores(0.99)for age and ethnicity-based bullying.A comparative analysis with state-of-the-art benchmarks,including Generative Pre-trained Transformer 2(GPT-2)and Text-to-Text Transfer Transformer(T5)models highlights BERT’s superiority in handling ambiguous language.This work advances cyberbullying detection by demonstrating how hybrid feature extraction and transformer models improve multi-class classification,offering a scalable solution for moderating nuanced harmful content.
文摘In the rapidly evolving landscape of natural language processing(NLP)and sentiment analysis,improving the accuracy and efficiency of sentiment classification models is crucial.This paper investigates the performance of two advanced models,the Large Language Model(LLM)LLaMA model and NLP BERT model,in the context of airline review sentiment analysis.Through fine-tuning,domain adaptation,and the application of few-shot learning,the study addresses the subtleties of sentiment expressions in airline-related text data.Employing predictive modeling and comparative analysis,the research evaluates the effectiveness of Large Language Model Meta AI(LLaMA)and Bidirectional Encoder Representations from Transformers(BERT)in capturing sentiment intricacies.Fine-tuning,including domain adaptation,enhances the models'performance in sentiment classification tasks.Additionally,the study explores the potential of few-shot learning to improve model generalization using minimal annotated data for targeted sentiment analysis.By conducting experiments on a diverse airline review dataset,the research quantifies the impact of fine-tuning,domain adaptation,and few-shot learning on model performance,providing valuable insights for industries aiming to predict recommendations and enhance customer satisfaction through a deeper understanding of sentiment in user-generated content(UGC).This research contributes to refining sentiment analysis models,ultimately fostering improved customer satisfaction in the airline industry.
基金supported by the National Natural Science Foundation of China under Grant No.U1736216。
文摘Traffic encryption techniques facilitate cyberattackers to hide their presence and activities.Traffic classification is an important method to prevent network threats.However,due to the tremendous traffic volume and limitations of computing,most existing traffic classification techniques are inapplicable to the high-speed network environment.In this paper,we propose a High-speed Encrypted Traffic Classification(HETC)method containing two stages.First,to efficiently detect whether traffic is encrypted,HETC focuses on randomly sampled short flows and extracts aggregation entropies with chi-square test features to measure the different patterns of the byte composition and distribution between encrypted and unencrypted flows.Second,HETC introduces binary features upon the previous features and performs fine-grained traffic classification by combining these payload features with a Random Forest model.The experimental results show that HETC can achieve a 94%F-measure in detecting encrypted flows and a 85%–93%F-measure in classifying fine-grained flows for a 1-KB flow-length dataset,outperforming the state-of-the-art comparison methods.Meanwhile,HETC does not need to wait for the end of the flow and can extract mass computing features.The average time for HETC to process each flow is only 2 or 16 ms,which is lower than the flow duration in most cases,making it a good candidate for high-speed traffic classification.
基金Prince Sattambin Abdulaziz University for funding this research work through the project number(PSAU/2025/R/1446).
文摘The early and precise identification of Alzheimer’s Disease(AD)continues to pose considerable clinical difficulty due to subtle structural alterations and overlapping symptoms across the disease phases.This study presents a novel Deformable Attention Vision Transformer(DA-ViT)architecture that integrates deformable Multi-Head Self-Attention(MHSA)with a Multi-Layer Perceptron(MLP)block for efficient classification of Alzheimer’s disease(AD)using Magnetic resonance imaging(MRI)scans.In contrast to traditional vision transformers,our deformable MHSA module preferentially concentrates on spatially pertinent patches through learned offset predictions,markedly diminishing processing demands while improving localized feature representation.DA-ViT contains only 0.93 million parameters,making it exceptionally suitable for implementation in resource-limited settings.We evaluate the model using a class-imbalanced Alzheimer’s MRI dataset comprising 6400 images across four categories,achieving a test accuracy of 80.31%,a macro F1-score of 0.80,and an area under the receiver operating characteristic curve(AUC)of 1.00 for the Mild Demented category.Thorough ablation studies validate the ideal configuration of transformer depth,headcount,and embedding dimensions.Moreover,comparison research indicates that DA-ViT surpasses state-of-theart pre-trained Convolutional Neural Network(CNN)models in terms of accuracy and parameter efficiency.
文摘Background:Early and accurate diagnosis of cataracts,which ranks among the leading preventable causes of blindness,is critical to securing positive outcomes for patients.Recently,eye image analyses have used deep learning(DL)approaches to automate cataract classification more precisely,leading to the development of the Multiscale Parallel Feature Aggregation Network with Attention Fusion(MPFAN-AF).Focused on improving a model’s performance,this approach applies multiscale feature extraction,parallel feature fusion,along with attention-based fusion to sharpen its focus on salient features,which are crucial in detecting cataracts.Methods:Coarse-level features are captured through the application of convolutional layers,and these features undergo refinement through layered kernels of varying sizes.Moreover,this method captures all the diverse representations of cataracts accurately by parallel feature aggregation.Utilizing the Cataract Eye Dataset available on Kaggle,containing 612 labelled images of eyes with and without cataracts proportionately(normal vs.pathological),this model was trained and tested.Results:Results using the proposed model reflect greater precision over traditional convolutional neural networks(CNNs)models,achieving a classification accuracy of 97.52%.Additionally,the model demonstrated exceptional performance in classification tasks.The ablation studies validated that all applications added value to the prediction process,particularly emphasizing the attention fusion module.Conclusion:The MPFAN-AF model demonstrates high efficiency together with interpretability because it shows promise as an integration solution for real-time mobile cataract detection screening systems.Standard performance indicators indicate that AI-based ophthalmology tools have a promising future for use in remote conditions that lack medical resources.
基金financially supported by the National Natural Science Foundation of China(No.52174297).
文摘The endpoint carbon content in the converter is critical for the quality of steel products,and accurately predicting this parameter is an effective way to reduce alloy consumption and improve smelting efficiency.However,most scholars currently focus on modifying methods to enhance model accuracy,while overlooking the extent to which input parameters influence accuracy.To address this issue,in this study,a prediction model for the endpoint carbon content in the converter was developed using factor analysis(FA)and support vector machine(SVM)optimized by improved particle swarm optimization(IPSO).Analysis of the factors influencing the endpoint carbon content during the converter smelting process led to the identification of 21 input parameters.Subsequently,FA was used to reduce the dimensionality of the data and applied to the prediction model.The results demonstrate that the performance of the FA-IPSO-SVM model surpasses several existing methods,such as twin support vector regression and support vector machine.The model achieves hit rates of 89.59%,96.21%,and 98.74%within error ranges of±0.01%,±0.015%,and±0.02%,respectively.Finally,based on the prediction results obtained by sequentially removing input parameters,the parameters were classified into high influence(5%-7%),medium influence(2%-5%),and low influence(0-2%)categories according to their varying degrees of impact on prediction accuracy.This classi-fication provides a reference for selecting input parameters in future prediction models for endpoint carbon content.
基金supported through Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2025R508)Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia。
文摘Unmanned Aerial Vehicles(UAVs)have become indispensable for intelligent traffic monitoring,particularly in low-light conditions,where traditional surveillance systems struggle.This study presents a novel deep learning-based framework for nighttime aerial vehicle detection and classification that addresses critical challenges of poor illumination,noise,and occlusions.Our pipeline integrates MSRCR enhancement with OPTICS segmentation to overcome low-light challenges,while YOLOv10 enables accurate vehicle localization.The framework employs GLOH and Dense-SIFT for discriminative feature extraction,optimized using the Whale Optimization Algorithm to enhance classification performance.A Swin Transformer-based classifier provides the final categorization,leveraging hierarchical attention mechanisms for robust performance.Extensive experimentation validates our approach,achieving detection mAP@0.5 scores of 91.5%(UAVDT)and 89.7%(VisDrone),alongside classification accuracies of 95.50%and 92.67%,respectively.These results outperform state-of-the-art methods by up to 5.10%in accuracy and 4.2%in mAP,demonstrating the framework’s effectiveness for real-time aerial surveillance and intelligent traffic management in challenging nighttime environments.
基金supported by the Natural Science Foundation of China(No.42301492)the National Key Research and Development Program(No.2022YFB3904200)+4 种基金the Open Fund of Key Laboratory of Urban Land Resources Monitoring and Simulation,Ministry of Natural Resources(No.KF-2022-07-014)the Natural Science Foundation of Hubei Province of China(No.2022CFB640)the Open Fund of Hubei Key Laboratory of Intelligent Vision Based Monitoring for Hydroelectric Engineering(No.2022SDSJ04)the Opening Fund of Key Laboratory of Geological Survey and Evaluation of Ministry of Education(No.GLAB 2023ZR01)the Fundamental Research Funds for the Central Universities.
文摘If progress is to be made toward improving geohazard management and emergency decision-making,then lessons need to be learned from past geohazard information.A geologic hazard report provides a useful and reliable source of information about the occurrence of an event,along with detailed information about the condition or factors of the geohazard.Analyzing such reports,however,can be a challenging process because these texts are often presented in unstructured long text formats,and contain rich specialized and detailed information.Automatically text classification is commonly used to mine disaster text data in open domains(e.g.,news and microblogs).But it has limitations to performing contextual long-distance dependencies and is insensitive to discourse order.These deficiencies are most obviously exposed in long text fields.Therefore,this paper uses the bidirectional encoder representations from Transformers(BERT),to model long text.Then,utilizing a softmax layer to automatically extract text features and classify geohazards without manual features.The latent Dirichlet allocation(LDA)model is used to examine the interdependencies that exist between causal variables to visualize geohazards.The proposed method is useful in enabling the machine-assisted interpretation of text-based geohazards.Moreover,it can help users visualize causes,processes,and other geohazards and assist decision-makers in emergency responses.
基金the National Natural Science Foundation of China, No. 81071001, 30600200the Natural Science Foundation of Hu-nan Province, No. 2006JJ30009
文摘Charcot-Marie-Tooth （CMT） disease is the most common inherited peripheral neuropathic disorder. CMT is clinically and genetically heterogeneous. To date, 27 genes associated with the disease have been cloned. The present study carried out clinical classification according to clinical, electrophysiological and pathological features, conducted inheritance classification according to inheritance patterns, and performed mutation analysis of 13 CMT disease genes （PMP22, CX32, HSPB1, MNF2, MPZ, HSPB8, GDAP1, NFL, EGR2, SIMPLE, RAB7, LMNA, MTMR2） in 57 Chinese probands with CMT. Five cases of AD-CMT1 and 13 cases of sporadic CMT1 were diagnosed as CMT1A; five cases of X-CMT1, one case of X-CMT2 and one case of sporadic CMT1 were diagnosed as CMTXl; four cases of AD-CMT2 were diagnosed as CMT2F; one case of AD-CMT2 and one case of sporadic CMT2 were diagnosed as CMT2A2; one case of AD-CMT2 was diagnosed as CMT2L; one case of AD-CMT2 was diagnosed as CMT2J; one case of AR-CMT1 was diagnosed as CMT4A. Among the 57 CMT probands, seven genotypes were determined among 34 patients, with a detection rate of 59.6%. The results indicated that the clinical classification and inheritance classification are indispensable for selecting potential disease genes for mutation detection, and for efficient molecular diagnosis.
