Search Results (36,529)

Addressing the critical challenge of signal degradation in biosensor cardiac monitoring, this paper introduces an efficient hybrid dilated–long short-term memory (LSTM) with residual learning (HDLR), which is a novel architecture engineered for a high-fidelity electrocardiogram (ECG) denoising signal. The proposed HDLR model synergistically integrates with dilated convolutions to expand the receptive field for multi-scale feature extraction. This is an LSTM-based backbone used to resolve long term temporal dependencies with residual learning paths to stabilize gradient flow and accelerate convergence. The proposed HDLR architecture integrates three core functional components with dilated convolutional layers utilized for local temporal feature extraction, where varying dilation rates expand the receptive field to capture both local waveform patterns and broader morphological structures without increasing computational complexity. Experimental results demonstrate a significant leap in performance, with the HDLR model achieving a mean squared error (MSE) of 0.002176, a signal-to-noise ratio (SNR) of 14.4420 dB, and a Matthews correlation coefficient (MCC) of 0.9822. Beyond quantitative metrics, the proposed HDLR architecture exhibits exceptional robustness in preserving cardiac morphology, specifically the P-wave and QRS complex of the ECG signal under stochastic noise conditions. These findings underscore the HDLR model’s potential as a backbone for next generation, real time diagnostic systems in intelligent healthcare. Full article

To address the limited extraction of local fault features, ineffective fusion of multi-scale fault information and interference from redundant noise in vibration-based fault diagnosis of rolling bearings and gears under complex operating conditions, a multi-scale gated feature refinement Vision Transformer (MGFR-ViT), is proposed. First, one-dimensional vibration signals are reconstructed as two-dimensional vibration matrices, making them compatible with patch embedding and Vision Transformer-based feature modeling. A locally enhanced Vision Transformer module is then developed by incorporating a local enhancement mechanism into the standard Vision Transformer architecture, thereby improving the extraction of locally fault-sensitive features while preserving global dependency modeling. Furthermore, a multi-scale gated feature refinement module is introduced to adaptively enhance fault-relevant information and suppress redundant features and noise through parallel multi-scale convolutions with different receptive fields, channel interaction, and gated weighting. Finally, global average pooling and a fully connected classifier are employed for fault classification. Experiments conducted on bearing and gear datasets demonstrated that MGFR-ViT achieved superior diagnostic performance and feature separability compared with several representative fault diagnosis models. Ablation studies further validated the effectiveness and complementarity of the proposed modules. These results indicate that MGFR-ViT provides an effective feature-learning framework for vibration-based fault diagnosis of rotating machinery. Full article

Background/Objectives: 22q11.2 microduplication syndrome is a rare genetic disorder characterized by the presence of one or two additional copies of a segment within the 22q11.2 region of chromosome 22. While much of the literature has focused on the deletion variant leading to DiGeorge syndrome, the duplication counterpart has gained increasing attention due to its clinical variability and under-recognition. This review aims to deliver new possibilities to genetic counseling that can be provided in prenatal and postnatal cases as the phenotype of 22q11.2 microduplication carriers cannot be fully predicted. Methods: In the present study, a total of 12 (5 prenatal and 7 postnatal) cases were diagnosed through array-CGH and combined with 679 (95 prenatal and 584 postnatal) cases reported in the literature. This review summarizes the published evidence available up to April 2025. Data on clinical presentations, genetic findings, diagnostic methodologies, and outcomes were extracted and analyzed. Results: The combination of our cases and the reported cases with 22q11.2 microduplication syndrome revealed a broad phenotypic spectrum. Common clinical features include neurodevelopmental disorders, and cardiac anomalies. Importantly, the syndrome exhibits variable expressivity and reduced penetrance, with more than 70% of the findings to be inherited by one of the parents. Conclusions: 22q11.2 microduplication syndrome presents a heterogeneous clinical picture with variable expressivity and incomplete penetrance, posing challenges in diagnosis and genetic counseling, particularly when predicting prenatal outcomes. Awareness of its diverse manifestations is crucial for clinicians to consider this syndrome in the differential diagnosis and to provide informed counseling. Full article

Background/Objectives: To develop a fast and interpretable multimodal framework for the automatic detection of cardiac abnormalities using electrocardiogram (ECG) and phonocardiogram (PCG) signals. Methods: A multimodal classification scheme was designed by combining ECG and PCG recordings. For each modality, tailored preprocessing and temporal and spectral feature extraction were applied. The resulting information was fused through a quality-aware strategy that prioritized more reliable signal segments. The explainability results showed that the model focused on physiologically meaningful regions, providing supportive interpretability for its predictions. Experiments were conducted on the PhysioNet/CinC 2016 heart sound dataset, including normal and pathological recordings, using 10-fold cross-validation. Results: The proposed method achieved a mean F1 score of 97.2%, an accuracy of 95.6%, a specificity of 88.6%, and a sensitivity of 97.7%. In addition, the lightweight preprocessing and fast feature extraction pipeline allowed the full 10-fold cross-validation procedure to be completed in only 66 s. Conclusions: The proposed ECG-PCG framework provides a fast, accurate, and interpretable solution for automated cardiac abnormality detection and appears well suited for real-time cardiac screening applications. Full article

Gait symmetry analysis has emerged as an important tool in rehabilitation engineering and neurological disorder assessment, as it provides clinically relevant indicators of mobility impairment and gait abnormalities. The proposed framework integrates gait symmetry variability, statistical gait features and Dynamic Time Warping (DTW)-based temporal alignment to enhance robustness against gait variations and irregular walking patterns. A hybrid feature vector comprising DTW similarity scores, the hybrid gait symmetry index (GSI), and statistical gait descriptors was employed to classify gait patterns into five categories: normal, slow, medium, fast, and abnormal. The system was implemented as a wearable edge-computing platform using an NI myRIO device equipped with a tri-axial Inertial Measurement Unit (IMU) mounted on the subject’s body. The onboard FPGA performs real-time signal preprocessing, GSI computation, feature extraction, constrained DTW matching, and gait classification using fixed-point streaming architectures and BRAM-based buffering. Meanwhile, the embedded ARM processor manages TCP/IP communication and transmits real-time gait information to a remote monitoring workstation via a WiFi interface for visualization and analysis. Operating at a clock frequency of 100 MHz, the complete architecture achieves an end-to-end processing latency of approximately 4 ms. The proposed FPGA-based implementation provides low-latency, energy-efficient, and real-time gait analysis, making it well suited for wearable rehabilitation systems, assistive healthcare devices, and continuous mobility monitoring applications. Full article

This article presents the use of autoencoders as part of a dimensionality-reduction method in the task of crime clustering. The study was conducted on a real-world crime dataset from the city of Chicago, based on publicly available police records. The initial data processing involved selecting and extracting variables, aggregating the selected variables, and converting crime categories and incident locations into contextual embeddings. The data prepared in this way was used to train various autoencoder architectures, including Vanilla, convolutional, denoising and variational models. The representations obtained from the latent layer of the encoder were then used as input data for clustering methods, such as k-means, Gaussian mixture model, and spectral clustering. The experimental results showed that the use of autoencoders in the clustering process enabled the identification of distinct groups of offences, with the best results (measured using the ARI and NMI metrics) obtained for Vanilla autoencoders combined with k-means and GMM, particularly with intermediate latent-space dimensions. The results confirm the potential of autoencoders as effective tools for dimensionality reduction and feature extraction in crime data analysis, as well as their usefulness in the exploratory analysis of complex urban data. Full article

The road surface adhesion coefficient is a key parameter affecting the dynamic control and driving safety of vehicles. Addressing the limitations of vehicle dynamics-based estimation methods under low-excitation conditions, this paper proposes a vision-perception-based CNN–Mamba hybrid regression network (CMHR-Net) for visual road friction potential estimation. By extracting road surface texture features from camera images, the proposed method predicts representative friction levels associated with different road conditions, providing prior information for vehicle active safety and control systems. First, ResNet18 is used to extract local texture features from road surface images. Second, a Mamba State Space Model is introduced to model long-distance dependencies between features, thereby enhancing global representation capabilities. Finally, the predicted adhesion coefficient value is output through a regression layer. Simultaneously, an adhesion coefficient mapping dataset based on road surface semantic attributes is constructed for model training and validation. Experimental results show that under various road surface conditions, including wet asphalt, waterlogged asphalt, waterlogged concrete, ice and snow, and joints, the proposed method significantly reduces the MAE (Mean Absolute Error) compared to the traditional CNN model. Specifically, under low adhesion conditions (μ ≈ 0.20), the error is reduced by approximately 56.9%. Furthermore, in complex variable conditions, the model significantly outperforms the traditional CNN model in both maximum error (Max Error) and mean absolute percentage error (MAPE). For example, under low adhesion conditions (μ ≈ 0.20), the MAPE decreases from 21.59% to 9.28%, and the maximum error decreases from 0.2025 to 0.0628, demonstrating superior stability and robustness. This method can achieve high-precision feedforward estimation of the adhesion coefficient without relying on vehicle dynamics excitation, providing effective support for feedforward control and active safety systems in vehicles. Full article

Global Navigation Satellite System Reflectometry (GNSS-R) platforms, such as the CYGNSS constellation, provide unprecedented spatial coverage for monitoring ionospheric scintillation via the $S_4$ index. However, the operational utility of GNSS-R for space weather is substantially degraded by surface-induced signal contamination when sharp land–water boundaries (coastlines) trigger massive, false-positive $S_4$ pseudo-scintillations that imitate true ionospheric plasma irregularities. In this study, a robust machine learning (ML) methodology to autonomously distinguish surface-induced reflections from true atmospheric volumetric scattering was proposed. Using 1 Hz Level 1 continuous Signal-to-Noise Ratio (SNR) time-series data, morphologic features (e.g., maximum amplitude, peak prominence, and standard deviation) were extracted to train a Random Forest (RF) classifier. The model achieves 98% accuracy in differentiating coastal boundaries from ionospheric scintillation, evaluated on a global dataset of over ~450,000 anomalous events. Moreover, a multi-sensor case study of the historic May 2024 G5 geomagnetic storm is presented to validate the geophysical fidelity of the filtered data. The ML-isolated CYGNSS anomalies demonstrate strong spatial correlation with COSMIC-2 Radio Occultation (RO) F2-peak electron density (NmF2) variations and ground-based Rate of TEC Index (ROTI) maps. Furthermore, temporal cross-validation with 1 Hz localized ground magnetometer data in Northwest Mexico reveals positive synchronization between CYGNSS scattering events and localized electrodynamic disturbances. Finally, the results demonstrate that ML-de-aliased GNSS-R data can reliably link the oceanic observational gaps inherent to ground-based networks, offering a powerful new tool for global space weather monitoring. Full article

The helicopter tail drive shaft is vulnerable to failure from projectile impacts during low-altitude flight. Stress wave-based inversion of impact parameters and quantitative damage assessment remain insufficiently explored. To address small-sample and nonlinear challenges, a framework based on stress wave characteristics and physics-guided hierarchical Gaussian process regression is proposed. Four key features, namely first-arrival wave trough amplitude, frequency standard deviation, ratio of low-frequency to high-frequency root mean square, and wavelet energy entropy, are extracted from transient signals to construct a hierarchical progressive architecture for damage mode discrimination, parameter inversion, and quantitative assessment. Perforation is identified using a wavelet energy entropy-based adaptive threshold. Incidence angle inversion is achieved by an adaptive composite kernel and Bayesian physical prior correction. Damage degree is assessed through residual learning guided by a physical prior surface mean function. Results show an incidence angle inversion root mean square error (RMSE) of 3.02°, with entry and exit hole equivalent failure area RMSEs of 13.32 mm² and 12.98 mm², respectively. The 95% prediction interval maintained reliable coverage across the validation samples. This framework provides a new method with both physical interpretability and uncertainty quantification for the assessment of impact damage in thin-walled tube structures. Full article

Particle image velocimetry (PIV), as a widely used technique for measuring flow velocity fields and characterizing flow behavior, is widely used across various fluid mechanics applications, such as aerodynamics and fluid mechanics. However, the existing PIV methods focus on spatial resolution, i.e., the ability to extract small-scale motions, while ignoring the situation of large-scale flow motion. In this paper, a novel deep learning framework called Transformer is introduced, which is based on the self-attention mechanism and can capture global information. First, in order to enhance the ability to sense location information, convolutional neural network (CNN) is utilized to extract features from original particle images. Second, self-attention mechanism is applied to capture global information based on a set of image features. Third, velocity fields of the particle images are estimated in the encoder-decoder network. Finally, a number of both synthetic and experimental particle images are used to verify the proposed method. The experimental results indicate that the proposed method is superior to cross-correlation and optical flow methods on measurement accuracy. Moreover, it has better performance than traditional deep learning-based PIV methods on large-scale flow motion estimation. Full article

Unmanned aerial vehicle (UAV) fault diagnosis is critical for ensuring flight safety, yet existing methods face challenges from class-imbalanced datasets and the lack of reliable uncertainty estimates. We address these issues with HCUNet, a hierarchical consistency-based uncertainty quantification network that integrates multi-level feature aggregation with principled uncertainty estimation. The framework extracts complementary representations through three sequential stages. First, an adaptive data-driven multi-scale convolution module captures shallow multi-frequency patterns. Second, a hierarchical context bridging LSTM encodes middle-level temporal dynamics. Third, a rotary position embedding enhanced attention mechanism extracts deep semantic features. These hierarchical representations are fused by a latent-aware semantic extraction and aggregation module, which employs cross-scale dilated convolutions and random pooling sampling for robust feature integration. A single-pass mechanism disentangles epistemic and aleatoric uncertainties via prediction disagreement and feature divergence, eliminating the need for ensemble models. The hierarchical consistency loss jointly optimizes multi-level classification accuracy, inter-level agreement, and confidence calibration. Experiments on the ALFA dataset demonstrate that HCUNet achieves 99.55% accuracy, 99.12% F1-score, and 0.0024 expected calibration error, outperforming the strongest baseline by 0.66% in accuracy and 64.2% in calibration error reduction. Full article

In the restoration of mural images with rich colour information and complex texture structures, existing techniques typically extract the spatial-domain features in the Red–Green–Blue (RGB) colour space. However, the three RGB channels are physically decoupled without unified perceptual colour correlation constraints, which often leads to noticeable colour deviation in damaged regions with large colour variations. In addition, restoring both high-frequency texture details and low-frequency global structures in a mixed-frequency spatial domain can create conflicts between frequencies, making it difficult to generate realistic high-frequency details. To address these issues, we propose the laboratory frequency network (LABFNet), a restoration network guided by the laboratory (LAB) colour space and frequency-domain constraints. Our model has two key improvements: (1) it incorporates colour parameters from the LAB space to model colour loss in murals, and (2) it decomposes the image into low- and high-frequency components and enforces frequency consistency during restoration. In the Dunhuang 20–40% mask-ratio setting, the Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) improved by 1.58% and 0.27%, respectively, while the Mean Absolute Error (MAE), Learned Perceptual Image Patch Similarity (LPIPS) and CIEDE2000 decreased by 5.87%, 6.5%, and 21.65%, respectively. Experimental results on benchmark datasets show that LABFNet reduces colour deviation and structural defects. Full article

Glioblastoma recurrence remains nearly inevitable despite maximal resection and adjuvant therapy, with most relapses occurring within or adjacent to the original tumor site. Conventional MRI underestimates tumor infiltration beyond contrast-enhancing margins, limiting preoperative identification of peritumoral regions at higher risk of recurrence. We developed a radiomics-based machine-learning framework to generate preoperative spatial recurrence risk maps from routine MRI. Preoperative T1-weighted contrast-enhanced and FLAIR images from 79 patients with glioblastoma were analyzed. The cohort was divided at the patient level into a training set (n = 63) and an independent test set (n = 16). Using a balanced spatial ROI sampling strategy, local radiomic features were extracted from tumor and peritumoral regions and linked to recurrence sites identified on follow-up MRI acquired after at least 12 months after surgery. A CatBoost classifier achieved an AUC of 0.743 and a recall of 0.856 on an independent test set. The framework also generated preoperative probability maps that showed qualitative spatial correspondence with observed recurrence locations. These findings indicate that radiomic patterns from standard preoperative MRI may contain spatially localized information associated with future relapse. The proposed approach supports the feasibility of preoperative spatial risk stratification and may provide useful information for surgical planning and subsequent treatment strategies. Full article

Small objects in video streams occupy a small proportion in the image; the texture and shape information they carry is limited, making it difficult to continuously track and identify. To solve this problem, a method for continuous tracking and recognition of small objects in the video stream based on YOLO and spatio-temporal context memory network is proposed. A backbone network based on the improved YOLOv8 model is introduced, and the multi-scale visual features of small objects are extracted at different levels of the video stream using the wavelet pooling module. A mixed attention module enhances the feature response in the spatially significant pixel regions, generating weighted multi-scale visual features. The neck network processes these weighted features through a spatio-temporal context memory network to extract multi-scale spatio-temporal features. Then, a bidirectional feature pyramid module fuses these multi-scale spatio-temporal features. The head network processes the fused features to output continuous recognition results for small objects. Experiments show that the proposed method successfully extracts the spatiotemporal features of small objects from video stream data sets dominated by small objects. Under different conditions of small object occlusion rates, this method achieves a success rate of continuous tracking and recognition of small objects over 0.93. Full article

Alzheimer’s disease (AD) is the most prevalent neurodegenerative disorder worldwide, affecting approximately 56.9 million people in 2021 and projected to reach 152 million by 2050. Its defining pathological features, amyloid-beta plaques and neurofibrillary tangles, accumulate for up to two decades before cognitive symptoms emerge, placing the preclinical and mild cognitive impairment (MCI) stages at the centre of the early detection problem. Despite this, current diagnostic practice in routine clinical settings remains unreliable, with post-mortem studies placing the specificity of clinical AD diagnosis between 44.3 and 70.8% even in specialist memory clinics. Machine learning (ML) and deep learning (DL) applied to neuroimaging and electrophysiological data have emerged as candidate tools for closing this diagnostic gap, yet whether the accuracy figures reported in published studies translate into clinically useful performance on independent data remains unresolved. This study presents a structured comparative review of machine learning and deep learning methods reported across four publicly available Alzheimer’s disease datasets, namely the Alzheimer’s Disease Neuroimaging Initiative (ADNI), the Open Access Series of Imaging Studies (OASIS), the OpenNeuro ds004504 electroencephalography (EEG) dataset, and the Kaggle Alzheimer’s magnetic resonance imaging (MRI) dataset. Thirteen model families are examined through the published literature rather than through new experiments, and for each model and dataset combination, the best accuracy reported in the source study is recorded alongside the model’s mathematical formulation. All performance figures reported in this abstract and throughout the paper are taken from the published studies reviewed, not from new experiments conducted by the authors. Across the reviewed studies, deep learning architectures pre-trained on ImageNet and fine-tuned on neuroimaging data are reported to produce the highest accuracy on MRI classification tasks. Residual Network (ResNet)-101 is reported at 98.21 percent on ADNI and 97.45 percent on OASIS, while the IncepRes fusion architecture reaches 98.35% on OASIS by combining multi-scale feature extraction from InceptionV3 with residual connectivity from ResNet152V2. Traditional machine learning classifiers remain competitive on tabular clinical and biomarker data, with Extreme Gradient Boosting (XGBoost) reaching 91% on ADNI multiclass features. Logistic Regression achieves 82 to 85% on binary ADNI classification and is the only classifier in this review that provides explicit per-feature prediction contributions without post hoc tooling. Gaussian Naïve Bayes achieves 80 to 83% on the same task. On the OpenNeuro EEG dataset, K-nearest neighbours (KNN) with singular value decomposition (SVD) entropy features achieves 91% binary accuracy, with feature engineering quality determining performance more reliably than classifier architecture. Eight principal findings emerge from the cross-dataset analysis. Binary classification consistently outperforms multiclass by 10 to 30% across all datasets, reflecting the genuine biological ambiguity of the mild cognitive impairment category. Dataset size and augmentation predict reported accuracy more reliably than model architecture. Ensemble methods outperform individual classifiers by 5 to 8% in nearly every imaging study. Deeper architectures can overfit small clinical cohorts. EEG models trail MRI models by approximately 10 to 15% on comparable binary tasks. Cross-dataset generalisation has not been systematically evaluated in most studies, and the few that have tested it report accuracy drops of 5 to 10% or more when models encounter data from different scanners or cohorts. Eight recurring limitations constrain the clinical utility of these findings. Small sample sizes and limited demographic diversity, severe class imbalance inflating raw accuracy metrics, poor cross-dataset generalisation driven by scanner heterogeneity, limited deep learning interpretability, the dominance of binary over multiclass tasks, the absence of longitudinal modelling despite available datasets, inadequate standardisation of preprocessing and evaluation protocols, and the signal-to-noise ratio constraints specific to EEG recordings of elderly patients collectively define the gap between benchmark performance and clinical readiness. Future work must prioritise multi-centre training cohorts, multimodal fusion architectures, longitudinal progression modelling, and standardised interpretability evaluation as non-optional requirements for any system intended for clinical deployment. Full article