Accurate determination of Pb²⁺ in environmental water remains challenging because background currents, signal overlap, and matrix-induced variations can compromise the reliability of conventional peak-current analysis. Herein, we developed a gold nanoparticle-modified glassy carbon electrode (AuNPs/GCE) integrated with square-wave voltammetry (SWV) and multivariate chemometric modelling for sensitive Pb²⁺ detection in water. The electrode provides a stable, highly conductive electroactive interface that promotes efficient electrochemical deposition and stripping of Pb²⁺ without additional surface functionalisation. We have successfully optimised preconcentration parameters (deposition potential: -1.1 V; deposition time: 120 s), amplitude: 20 mV, and a scan rate 100mV/sec at which the method exhibits a linear response over the concentration range of 10–100 ppb, with a sensitivity of 0.50 µA ppb⁻¹ and a limit of detection (LOD) of 1.58 ppb. Unlike conventional approaches that quantify Pb²⁺ based solely on stripping peak currents, the proposed methodology leverages the full voltammetric fingerprint via multivariate chemometric modelling. Principal Component Analysis (PCA) demonstrated concentration-dependent variations in the voltammetric responses. The optimised Partial Least Squares Regression (PLSR) model exhibited a correlation coefficient of 0.9984, an R² of 0.9968, a calibration slope of 0.9968, an intercept of 0.1757 ppb, a prediction bias of 0.0000 ppb, and an RMSEC of 1.62 ppb. Under full cross-validation, the model yielded an R² of 0.9951 and an RMSECV of 2.38 ppb, demonstrating consistent predictive performance. The modified electrode further demonstrated good repeatability, satisfactory selectivity toward common coexisting ions, and reliable analytical performance in spiked tap water samples. The integration of SWV with multivariate chemometric modelling of the complete voltammetric fingerprint provides a simple, scalable, and computationally enhanced analytical framework for sensitive Pb²⁺ monitoring and offers a practical strategy for electrochemical analysis of complex environmental water samples.
Conventional fire detectors have difficulty penetrating obstacles for effective fire identification and localization. To overcome this limitation, a fire microwave radiation detection system was developed to enable reliable fire detection under obstructed conditions. The system employs a gimbal-mounted microwave radiometer that performs periodic scanning of the target area to collect microwave radiation signals and generate radiometric images. An edge detection algorithm combining Otsu’s thresholding and gradient methods was designed to locate the fire source center, while reflection analysis was integrated to eliminate localization errors caused by reflective interference. Experimental results show that the proposed method can detect fire sources in the presence of the tested non-metallic obstacle and can effectively suppress reflection-induced interference during source localization. In the localization experiments, the maximum horizontal and pitch angular errors were 1.88° and 1.8°, respectively, and the algorithmic processing time was less than 0.5 s. These findings demonstrate the potential of microwave radiometric imaging combined with the proposed localization algorithm for fire-source localization under obstruction and reflection interference, providing a promising complementary approach to conventional fire-detection technologies.
A dual-material-gate (DMG) MOS-HEMT biosensor is proposed for detection of breast cancer cells. The structure of the nanocavity beneath the gate acts as the sensing region wherein the dielectric properties of the target cells affect the response of the device. The experimentally reported dielectric constants of normal MCF-10A and malignant HS578T, MDA-MB-231, MCF-7, and T47D breast cancer cell lines are taken as the analyte conditions. The proposed design of DMG MOS-HEMT has been compared with that of a conventional single-material-gate (SMG) MOS-HEMT design. Since, the gate materials in proposed device have different work-function, a channel potential gradient will be established. Thus, an extra electrostatic degree of freedom exists in controlling the sensing channel. The 96.1
Near-field imaging techniques enable nanoscale characterization by probing localized probe-sample interactions that are inaccessible to conventional far-field imaging due to the diffraction limit. Optical, acoustic, and microwave near-field systems have emerged as powerful platforms for high-resolution sensing of physical, chemical, and electromagnetic properties in materials science, biomedical research, and semiconductor device characterization. However, conventional near-field imaging often suffers from slow point-by-point scanning, complex probe-sample interaction, and low signal-to-noise ratios, which restrict imaging efficiency and quantitative sensing performance. Recent advances in artificial intelligence (AI) provide new opportunities to overcome these limitations. AI-based approaches facilitate noise suppression, artifact correction, adaptive system control, and efficient inversion of physical parameters from near-field signals. By combining data-driven learning with physical constraints, AI-assisted near-field imaging systems can achieve improved stability, faster data acquisition, and enhanced quantitative sensing capabilities. This review discusses the technical principles and imaging modalities of near-field techniques, with a particular focus on AI-enabled intelligent sensing and safety monitoring in materials science, biomedicine, semiconductor and device characterization, and industrial applications, with the aim of providing more comprehensive technical support for the exploration of the microscopic world.
Automated gastrointestinal (GI) image classification remains challenging because of subtle morphological variation, high inter-class similarity, and the computational constraints of clinical edge devices. We present CA-MobileNetV4, a parameter-efficient teacher-student framework that combines a MobileNetV4-Small student, Channel-Prior Convolutional Attention (CPCA), a ConvNeXt-Tiny teacher enhanced with ACmix and Efficient Channel Attention (ECA), and Class Attention Transfer-based Knowledge Distillation (CAT-KD). CPCA is inserted in Layers 3 and 4 of the student to combine channel priors with multi-scale strip convolutions. During distillation, the pre-trained teacher is frozen; class activation maps (CAMs) from four semantic stages are projected to the class dimension, adaptively pooled to matching spatial sizes, L2-normalized, and aligned across all classes. Focal Loss is used as the classification term, and the CAT loss weight is set to β = 10 after evaluation of 5, 10, 15. Experiments use six Kvasir-v2 classes and eleven Lower-GI Hyper-Kvasir classes with stratified 70
