Sensors最新文献

Optical defect detection based on bright-field imaging is currently one of the most widely applied inspection techniques in wafer fabrication. However, particle defects on the surface of patterned wafers are often small in size. Under bright-field optical imaging conditions, defect signals are easily overwhelmed by complex background textures and noise, seriously affecting the detectability and positioning accuracy of defects. To address this issue, this paper proposes BWD-DETR, a detection framework tailored for wafer surface defects under bright-field imaging. Based on the RT-DETR baseline, this framework integrates a wavelet backbone, an SMFI module, and a CAS-Fusion module, achieving an AP₅₀ of 96.56% and an AP_50:95 of 54.94% in bright-field wafer defect detection, with improvements of 1.64% and 2.17% over the baseline, respectively. The proposed method can effectively enhance the detection capability for sub-micron defects on the wafer surface.

Recent advances in deep learning (DL) have enabled the integration of diverse biomedical data for disease prediction and risk stratification. Building on this progress, the overall objective of this study was to develop and evaluate a multimodal DL framework for robust multi-label classification (MLC) of major comorbidities in patients with obstructive sleep apnea (OSA) using physiological time series signals and clinical data. This study proposes a robust framework for multi-label classification (MLC) of comorbidities in patients with OSA using diverse physiological and clinical data sources. We conducted a retrospective observational study including a convenience sample of 144 patients referred for overnight polysomnography at the Sleep Medicine Center (SleepLab Split), University Hospital Centre Split (KBC Split), Split, Croatia. Patients were selected based on predefined inclusion criteria and data availability. A one-dimensional Convolutional Neural Network (1D-CNN) was developed to process and fuse time series signals, oxygen saturation (SpO2), derived SpO2 features, and nasal airflow (FP0), with demographic and physiological parameters, enabling the identification of key comorbidities such as arterial hypertension, diabetes mellitus, and asthma/COPD. The instruments included polysomnography-derived signals (SpO₂ and FP0 airflow) and structured demographic/physiological parameters. Signals were preprocessed and used as inputs to the proposed fusion model. The proposed model was trained and fine-tuned using the Optuna hyperparameter optimization framework, addressing class imbalance through weighted loss adjustments. Its performance was comprehensively assessed using multi-label evaluation metrics, including macro/micro F1-score, AUC-ROC, AUC-PR, subset and partial accuracy, Hamming loss, and multi-label confusion matrix (MLCM). The study protocol was approved by the Ethics Committee of the School of Medicine, University of Split (Approval No. 003-08/23-03/0015, Date: 17 October 2023). The 1D-CNN achieved superior predictive performance compared to traditional machine learning (ML) classifiers with macro AUC-ROC = 0.731 and AUC-PR = 0.750. The model demonstrated consistent behavior across age, gender, and BMI groups, indicating strong generalization and minimal demographic bias. In conclusion, the results confirm that SpO2 and airflow signals inherently encode comorbidity-specific physiological patterns, enabling efficient and scalable screening of OSA-related comorbidities without the need for full polysomnography. Although the study is limited by data set size, it provides a methodological basis for the application of multi-label DL models in clinical decision support systems. Future research should focus on the expansion of multi-center datasets, thereby improving model interpretability and potential clinical adoption.

Deep learning has achieved impressive progress in automated electrocardiogram (ECG) analysis, yet its performance still relies heavily on large-scale labeled datasets. As ECG annotation requires cardiologists, this process is costly and time-consuming, limiting its scalability in clinical practice. Contrastive learning offers a promising alternative by enabling the extraction of generalizable representations from unlabeled ECG data. In this study, we propose Anatomy-Aware Contrastive Learning for ECG (ACL-ECG), a self-supervised method that incorporates cardiac anatomical relationships into contrastive learning. ACL-ECG employs a physiology-aware augmentation strategy to generate rhythm-preserving augmented views, including random scale cropping, cardiac-cycle masking, and temporal shifting. Furthermore, ECG leads are grouped into four anatomically meaningful regions-anterior, inferior, septal, and lateral-and region-level contrastive objectives are introduced to promote intra-region consistency while enhancing inter-region discriminability. Extensive evaluations of downstream tasks demonstrate that ACL-ECG consistently outperforms state-of-the-art contrastive baselines under linear probing, achieving improvements of up to 1.29% in the area under the receiver operating characteristic curve (AUROC) and 3.57% in the area under the precision-recall curve (AUPRC). Moreover, when fine-tuned using only 10% of labeled data, ACL-ECG attains a performance comparable to fully supervised training, effectively reducing annotation requirements by approximately 5∼8×. Ablation studies further confirm the contributions of both the physiology-aware augmentation strategy and the anatomy-aware contrastive objective. Overall, ACL-ECG enhances representation quality without increasing annotation burden, and provides a promising and anatomy-informed foundation for self-supervised ECG analysis in label-scarce settings.

The Global Navigation Satellite System (GNSS) is commonly used for outdoor positioning. However, its effectiveness diminishes in urban canyons and indoor environments attributed to signal blockage. This study aims to explore the potential of GNSS signals penetrating indoor spaces through windows and to enhance indoor positioning with Three-Dimensional Mapping-Aided (3DMA) GNSS, a concept generally applied outdoors. The research employs a 3D model of a corridor with manually labeled window locations to predict satellite visibility within indoor areas. The study integrates Pedestrian Dead Reckoning (PDR) with an indoor Shadow-matching (I-SM) technique, utilizing an Extended Kalman Filter (EKF) to improve positioning accuracy. One of the findings indicates that the proposed method significantly enhances positioning performance and its availability, achieving a root mean square error (RMSE) that is 2 m better than using PDR alone or single epoch I-SM. The study concludes that integrating GNSS with I-SM technique and PDR can optimize an indoor positioning solution and highlights the potential for improved navigation solutions in complex urban environments.

Human activity recognition (HAR) based on inertial measurement units (IMUs) embedded in wearable devices has gained increasing relevance in healthcare, wellness, and fitness monitoring. However, most existing classification methods assume a closed-set setting, where all activity classes need to be defined during training, which limits their applicability in real-world environments where unseen or unexpected activities are present. To overcome this limitation, we adopt an open-set recognition (OSR) framework that requires minimal changes to the HAR classifiers traditionally employed for this purpose. We also provide an extensive empirical evaluation based on a leave-one-activity-out validation protocol applied to two datasets with IMU signals acquired from smart eyewear: a proprietary dataset and the publicly available UCA-EHAR dataset. A lightweight one-dimensional convolutional neural network was trained to classify six-axis IMU data across common activities. We assess open-set HAR performance using several methods requiring limited computational overhead and operating in the logit space, including maximum logit, Gaussian Mixture Models, Kernel Density Estimation, OpenMax, and Nearest Neighbor Distance Ratio. Robust identification of unknown activities was achieved, with area under the ROC curve > 0.8. These findings highlight the potential of low-complexity open-set approaches for real-time HAR on resource-constrained wearable platforms, supporting the development of adaptive and reliable sensor-based recognition systems for real-world use.

This study evaluates the feasibility of estimating stiffness-related parameters and pulse wave velocity (PWV) in a controlled in vitro circulatory setup using artificial silicone vessels with systematically varied Shore A hardness and wall thickness. From synchronized pressure and capacitive waveforms, fiducial points and engineered features are extracted, together with pump settings (stroke volume and heart rate). A Sugeno-type adaptive neuro-fuzzy inference system (ANFIS) is used for hardness-level prediction and benchmarked against linear regression and contemporary machine-learning/deep-learning baselines using stratified cross-validation. PWV estimates derived via hardness-to-elasticity conversion models and the Moens-Korteweg formulation are evaluated against a reference PWV obtained within the same experimental configuration. Under these controlled conditions, the proposed pipeline shows strong agreement with reference labels and measurements. The results should be interpreted as an in vitro validation step; translation to biological tissues or in vivo data will require external validation, calibration of material-property mapping, and robustness testing under physiological variability and measurement noise.

The proliferation of massive antenna arrays and the consequent intensification of near-field effects with 6G necessitate addressing critical security challenges in near-field communication environments. This paper presents a novel artificial noise-aided spatial and directional modulation (SDMN-AN) framework, specifically tailored for secure near-field communications. The proposed system integrates legitimate receiver indices, modulation symbols, and artificial noise (AN) confined to the null space of legitimate channels, thereby enhancing both spectral efficiency and communication security. Two precoding strategies-maximum-ratio transmission (MRT) and zero-forcing (ZF)-are investigated, offering trade-offs between hardware complexity and detection overhead. Analytical derivations of bit error rate (BER) bounds, corroborated by simulation results, underscore the superiority of the SDMN-AN framework in mitigating eavesdropping threats while significantly improving spectral efficiency, positioning it as a compelling solution for next-generation secure wireless networks.

A chemiresistive nitric oxide (NO) gas sensor based on Pt/WO₃ co-decorated carbon nanofibers (CNFs) was fabricated using a simple and scalable electrospinning process. This sensor demonstrates high-ppb-level NO detection at room temperature (25 °C), with an experimentally demonstrated detection limit of 100 ppb. It exhibits rapid response, good signal repeatability, excellent batch-to-batch reproducibility, and high selectivity toward NO. Compared with previously reported NO sensors, this work highlights the integration of Pt and WO₃ within a conductive CNF network, enabling room-temperature NO detection down to 100 ppb using a simple chemiresistive architecture. In addition, preliminary sensing tests were conducted using dried simulated breath samples prepared by introducing exogenous NO into exhaled breath from healthy volunteers, demonstrating the sensor's capability to resolve different NO levels in a complex breath-related background. Owing to its reliable performance and cost-effective fabrication, the sensor holds potential as a NO sensing platform, providing a materials-level basis for future breath NO analysis and other related applications.

Wildfires are environmental hazards with severe ecological, social, and economic impacts. Wildfires devastate ecosystems, communities, and economies worldwide, with rising frequency and intensity driven by climate change, human activity, and environmental shifts. Analyzing wildfire insights such as detection, predictive patterns, and risk assessment enables proactive response and long-term prevention. However, most of the existing approaches have been focused on isolated processing of data, making it challenging to orchestrate cross-modal reasoning and transparency. This study proposed a novel orchestrator-based multi-agent system (MAS), with the aim of transforming multimodal environmental data into actionable intelligence for decision making. We designed a framework to utilize Large Multimodal Models (LMMs) augmented by structured prompt engineering and specialized Retrieval-Augmented Generation (RAG) pipelines to enable transparent and context-aware reasoning, providing a cutting-edge Visual Question Answering (VQA) system. It ingests diverse inputs like satellite imagery, sensor readings, weather data, and ground footage and then answers user queries. Validated by several public datasets, the system achieved a precision of 0.797 and an F1-score of 0.736. Thus, powered by Agentic AI, the proposed, human-centric solution for wildfire management, empowers firefighters, governments, and researchers to mitigate threats effectively.

Background: The Balance Error Scoring System (BESS) is the most practiced static postural balance assessment tool, which relies on visual observation, and has been adopted as the gold standard in the clinic and field. However, the BESS can lead to missed and inaccurate diagnoses-because of its low inter-rater reliability and limited sensitivity-by missing subtle balance deficits, particularly in the athletic population. Smartphone technology using motion sensors may act as an alternative option for providing quantitative feedback to healthcare clinicians when performing balance assessments. The primary aim of this study was to explore the discriminative validity of an alternative novel smartphone-based cloud system to measure balance remotely in soccer athletes with and without hip pain.

Methods: This is an exploratory cross-sectional study. A total of 64 Australian soccer athletes (128 hips, 28% females) between 18 and 40 years completed single and tandem stance balance tests that were scored using the modified BESS test and quantified using the smartphone device attached to their lower back. An Exploratory Factor Analysis (EFA) and a Clustered Receiver Operating Characteristic (ROC) using an Area Under the Curve (AUC) were used to explore the discriminative validity between the smartphone sensor system and the modified BESS test. A Linear Mixed-Effects Analysis of Covariance (ANCOVA) was used to determine any statistical differences in static balance measures between individuals with and without hip-related pain.

Results: EFA revealed that the first factor primarily captured variance related to smartphone measurements, while the second factor was associated with modified BESS test scores. The ROC and the AUC showed that the smartphone sway measurements in the anterior-posterior and mediolateral directions during single-leg stance had an acceptable to excellent level of accuracy in distinguishing between individuals with and without hip-related pain (AUC = 0.72-0.80). Linear Mixed-Effects ANCOVA analysis found that individuals with hip-related pain had significantly less single-leg balance variability and magnitude in the anteroposterior and mediolateral directions compared to individuals without hip-related pain (p < 0.05).

Conclusion: Due to the ability of smartphone technology to discriminate between individuals with and without hip-related pain during single-leg static balance tasks, it is recommended to use the technology in addition to the modified BESS test to optimise a clinician-led assessment and to further guide clinical balance decision-making. While the study supports smartphone technology as a method to assess static balance, its use in measuring balance during dynamic movements needs further research.