Ceftriaxone has a similar effect on astrocytic and neuronal GLT-1 distribution after mild traumatic brain injury.

ABSTRACT Rock glaciers are an indispensable component of mountain cryosphere depicted as a promising future water reserve under ongoing global warming scenario. Conventionally, rock glaciers have been studied using high-resolution datasets employing geomorphic and kinematic-based approaches. The newly ventured domain of automatic detection of rock glaciers proved to be challenging and requiring competent skill and computation. The purpose of this study is to encourage the early career researchers to explore simpler techniques to detect and distinguish rock glaciers automatically, using free datasets. The present study provides a very simple semi-automatic method to differentiate rock glaciers developed in the complex mountain settings exploring widely accessible medium resolution optical datasets such as Sentinel 2A. The topographical peculiarity of rock glaciers were infused using freely obtained Advanced Land observing Satellite (ALOS) Digital Elevation Models (DEM). The visible and infrared bands of Sentinel 2A with hillslope and slope parameters were infused following some elementary image processing steps such as band ratio, principal component analysis (PCA) and band enhancement techniques to detect and distinguish rock glaciers semi-automatically.

Investigation of heat transfer performance of asphalt pavements at different compaction levels based on CT reconstruction

A novel machine learning assisted weight measurement system based on bending loss of optical fiber by using image processing

• GIS-driven aerial remote sensing supports large-scale power distribution monitoring. • Satellite images enable detection of clandestine areas linked to energy theft. • Vegetation management can benefit from GIS-driven aerial remote sensing. • Automatic estimation of rooftop PV rated capacity to update utility database. Solutions enabled by recent technological advancements and the increased availability of free geospatial images and open-source tools have demonstrated the potential to enhance tasks in various areas; however, electric power delivery remains insufficiently explored. This paper presents a proof-of-concept study exploring novel integrations of satellite imagery and geographic information systems (GIS) data to support three key tasks of distribution utilities worldwide: identifying clandestine connections to the system (electricity theft), mapping vegetation encroachment that poses risks to the network, and detecting and estimating the installed capacity of rooftop photovoltaic systems for automatically feeding or updating the utility database. The solutions rely on open-source tools, including artificial intelligence, image processing, and color segmentation, and are validated using real data from a Brazilian utility. The results demonstrate that the integration of GIS-driven and image-based aerial remote sensing techniques offers scalable and cost-efficient alternatives to conventional inspection methods.

Beam teleportation for precision dose control in scanning transmission electron microscopy.

ASIC design of trimmed weighted mean filter for image processing applications: An analog implementation

Large language models (LLMs), such as ChatGPT, are becoming increasingly prevalent, particularly in medical education and clinical assessments. Previous LLMs were seen to perform at the level of a first-year resident on the 2022 Orthopaedic In-Training Examination (OITE). With exponential advances in LLMs over the past 3 years, the true capabilities of these models remain unexplored. In addition, the addition of image processing further increases their clinical applicability. The purpose of this study was to evaluate the performance of six LLMs on the 2024 OITE. Six LLMs were evaluated in this study: ChatGPT (GPT-4o), Gemini 2.0 Flash, Grok 3, Mistral Large 2.7, DeepSeek R1, and Llama. ChatGPT, Gemini, Grok, and Mistral could evaluate images and text while DeepSeek and Llama were limited to text. Accuracy, image interpretation, and logical consistency were assessed in 203 multiple-choice questions, stratified by difficulty and type of the question. Statistical analyses involved chi-square tests, Fisher exact tests, z-tests, and Cohen κ tests. ChatGPT performed with the highest accuracy (74.9%), followed by DeepSeek, Llama, Grok, Mistral, and Gemini. ChatGPT also led in logical consistency (72.4%) and image interpretation (73.8%). Logical consistency strongly correlated with accuracy and correctness ( P < 0.00001). As difficulty increased, performance declined across all models. ChatGPT consistently scored the highest in terms of accuracy across all metrics while also maintaining reasoning quality. Compared with resident averages, ChatGPT performed at a postgraduate year five level which indicates its potential for integration into orthopaedic clinics, electronic medical records, and surgical planning. Further development models would allow for better performance on difficult questions and creating orthopaedic focused models could enhance these results.

During acute ischemic stroke, cerebral tissue undergoes different stages of ischemic damage and evolves towards irreversible injury at a varying pace depending on local perfusion and metabolic factors. This complex ischemic pathological process represents a dynamic continuum that has been historically conceptualized as a binary ischemic core-penumbra model. Although this simplification has proven useful for explaining the evolution of tissue damage in acute stroke, important nuances with clinical implications might be underappreciated. In this review, we critically appraise the pathophysiology and conventional clinical concepts adopted to explain infarct evolution in the early phases of ischemic stroke. We discuss recent mounting evidence that challenges the traditional compartmentalization of the ischemic core, penumbra, and oligemia, calling for more nuanced pathophysiological tissue concepts. For example, clinical benefits and the harmful hemorrhagic transformation associated with reperfusion therapies are observed across a spectrum of core volumes, challenging the deterministic assumptions of the core-penumbra hypothesis. Automated image processing systems reinforce this simplification of stroke pathophysiology, leading to misinterpretation of the range of truth in human imaging. We propose a modified definition of the core-penumbra-oligemia continuum that includes 6 levels of ischemic progression and their corresponding clinical implications: (1) benign oligemia, (2) vulnerable oligemia, (3) durable penumbra, (4) critical penumbra, (5) nonleaky core, and (6) leaky core. This more granular classification could better reflect the continuum of pathological ischemic changes and vulnerability. The proposed 6 levels can provide a framework for future neuroimaging efforts to better understand tissue fate and infarct evolution in ischemic stroke, ultimately informing treatment decision-making and refining targeting for new therapeutic approaches.

Lining surface defects adversely affect boiler operation and efficiency, and pose safety risks. The existing crack detection methods for refractory materials lack intelligence and quantitative precision. To address these limitations, well-designed crack detection methods are required for defect identification and maintenance. An explainable AI (XAI) model was developed to detect lining surface defects in a 700 MW circulating fluidized bed (CFB) boiler in Yunnan, China. The hybrid model combines deep learning (i.e., faster region-based convolutional neural network, FRCNN) and image processing to automatically identify and classify defects. To enhance explainability, FRCNN and Vision Transformer (ViT) are used to highlight the contribution of each feature to defect identification. The experimental results show that the hybrid model has high precision and recall rate in defect detection. A relative improvement of 28.01% in overall detection performance is achieved by the proposed ViT-FRCNN hybrid model, in comparison with the single detection method FRCNN. Meanwhile, an overall detection precision of over 95% is attained by the same hybrid model. This work supports CFB boiler operation and maintenance and offers innovative defect detection approaches for related fields. Future work will focus on lightweight Transformer adaptations to enhance computational efficiency for real-time industrial deployment. • Innovative XAI model for detecting surface defects in 700 MW CFB boilers. • Integration of FRCNN and ViT for enhanced defect identification. • Improved detection efficiency by 28.01% and accuracy over 95%. • Provides visual explanations for defect identification results. • Offers new ideas for defect detection in related industrial fields.

Mapping mosquito flight dynamics and directional responses: A scalable deep learning model for behavioural research.

This study evaluates the repeatability of a 3D simultaneous H/ Na MR fingerprinting (MRF) sequence in knee cartilage of healthy volunteers. Eight healthy volunteers underwent four knee scans each with 3D simultaneous H/ Na MRF at 7 T. Proton density (PD), tissue sodium concentration (TSC), and H and Na relaxation time maps were acquired over two visits with two consecutive acquisitions at both visits. Mean values and standard deviations of all MRF metrics were measured in three knee cartilage regions: patellar, femorotibial medial, and femorotibial lateral. Image processing included H and Na MRF dictionary matching, B correction, TSC and PD quantification, and image registration between the scans. Repeatability was assessed using the coefficient of variation (CV) and intraclass correlation coefficient (ICC) for all measurements and Bland-Altman plots to compare intraday and interday measurement differences. Mean TSC values over all subjects and cartilage regions were 162 29 mM, with a CV of 12% 1%. Mean Na T (30 2 ms) and T (13 3 ms) values were relatively consistent (CV = 6%-19%), while T showed greater variability (1.62 1.60 ms, CV = 54% 11%). For H, mean PD was 0.87 0.24 (CV = 20% 8%), mean T was 1114 168 ms (CV = 11% 5%), and mean T was 36 15 (CV = 20% 7%). ICC values suggested low-to-moderate discrimination power, with highest values observed for H T and lowest values for Na relaxation times. The Bland-Altman plots suggest similar intraday and interday differences. This study shows that we can acquire quantitative H and Na MRF maps simultaneously in knee cartilage at 7 T. Repeatability was the highest for TSC and Na T . ICC was the highest for PD and H T .

Abstract. Emission measurement of on-road vehicles in traffic is an important step for air pollution control and, thus, the reduction of negative effects on public health. Remote emission sensing (RES) is a state-of-the-art technology to detect high emitters by monitoring thousands of vehicles in traffic continuously. State-of-the-art (SOTA) RES systems use optical techniques to measure the ratio of specific pollutants to CO2 in vehicle exhaust plumes in order to determine emission factors. Highly accurate SOTA systems use laser absorption spectroscopy for measurement of the pollutant ratio in vehicle exhaust plumes. To obtain the absolute concentration of the single pollutants in the exhaust plume, the absorption path length must be known. In this work we present a 3D gas schlieren imaging sensor (GSIS) system which allows the geometrical reconstruction of 3D density fields of vehicle exhaust plumes in RES applications. Thus, it allows us to obtain the vehicle exhaust plume size and thereby enables estimation of the absorption path length from any direction. Furthermore, it is possible to determine where the laser intersects with the exhaust plume and, thus, to assess if the measurement is valid. The 3D-GSIS system consists of an array of low-cost digital cameras operating in the range of visible light. By means of advanced image processing and tomographic reconstruction techniques, the 3D displacement and density fields of vehicle exhaust plumes can be reconstructed. For validation, we characterized the 3D-GSIS system in the lab using hot air and CO2 plumes. Moreover, the 3D density fields of on-road passing vehicles are estimated and reconstructed using the 3D-GSIS system. The 3D-GSIS system is to be combined with an advanced RES system to measure the direct concentration of pollutants in vehicle exhaust plumes in the future.

A hybrid system integrating deep learning and computer vision for automated blink monitoring and tear film break-up pattern classification.

Component substitution and multiresolution analysis on hyperspectral microwave microscopy data sets.

Fourth-generation synchrotron radiation sources based on diffraction-limited storage rings impose stringent requirements on beamline alignment precision, diagnostic reliability, and real-time monitoring under high-brightness operating conditions. To address these requirements, an integrated fluorescent-target imaging and real-time beam diagnostic platform has been developed at the High Energy Photon Source (HEPS). The system incorporates a radiation-tolerant and ultra-high-vacuum-compatible mechanical assembly with diamond and YAG:Ce scintillators for different heat-load conditions. A distributed multi-camera architecture implemented within the Experimental Physics and Industrial Control System (EPICS) areaDetector framework enables synchronized multi-angle monitoring of beam profiles. Key beam parameters are extracted through online image processing and published as EPICS process variables for control-system integration. To support high-throughput diagnostic data handling, the platform further integrates the Mamba Data Worker framework, enabling coordinated multi-target acquisition, HDF5 storage, and automated metadata ingestion into a dedicated HEPS beamline alignment database. Representative deployment demonstrated stable synchronized operation of 13 cameras, with online analysis completed in less than 20 ms and end-to-end latency below 50 ms. These results establish a practical and scalable framework for beamline diagnostics, alignment support, and data-driven optimization atHEPS.

On-orbit image processing technology for intelligent remote sensing satellites: Progress, challenges, and opportunities

This study establishes a machine learning-enhanced computational framework to decode the pore-permeability constitutive relationship in fuel-related P(AA₀.₃-r-St₀.₇)₁₈₅ random copolymers. This copolymer exhibits good oil/fast resistance and controllable nanopore structure, making it ideal for fuel separation membranes, fuel filtration, and gas reservoir permeation in fuel-related energy systems. Through multi-scale modeling combining (i) Materials Studio-based conformational sampling with Markov chain Monte Carlo optimization, and (ii) CNN-assisted grayscale analysis of simulated pore structure images, we quantitatively correlate nanoscale porosity features (Connolly surface area, structural porosity from image processing, and specific surface area) with macroscopic transport properties. A simple CNN is used as an auxiliary image-processing tool to efficiently extract porosity from grayscale images. The Kozeny-Carman equation reveals a bifurcated linear correlation, exhibiting an inflection point at a porosity of 0.131, which corresponds to 23.2% cell-face porosity in the optimized models. Microstructural analysis indicates this critical transition stems from pore network reorganization, where permeability evolution shifts from rapid growth (< 0.131) to gradual increase (> 0.131). Validated against experimental trends, this hybrid computational approach demonstrates superior robustness to empirical methods by simultaneously capturing (a) molecular-level conformational dynamics through MS simulations, and (b) mesoscale porosity-permeability coupling via image-based machine learning. The framework provides a predictive tool for designing functional copolymer membranes for fuel processing and energy storage applications, with potential extensions to other porous material systems through transfer learning of the established microstructure-transport.

Retinotopic tuning of neural populations is a key organizing principle of human visual cortex. However, state-of-the-art models that predict neural recordings based on task-optimized Convolutional Neural Networks (CNNs) do not take this retinotopic organization into account. Furthermore, while retinotopic tuning in visual cortex has been studied extensively using functional magnetic resonance imaging, the temporal dynamics of processing information from distinct parts of the visual field are less well understood. Here, we reveal distinct temporal profiles for foveal and peripheral visual information processing by implementing multiple spatial sampling strategies on feature maps of CNNs into encoding models that predict human electroencephalography (EEG) responses. Using large, high-quality natural scene images, we show that processing of peripheral information precedes that of foveally sampled information. This temporal difference is best modeled when applying a differential spatial transform to CNN feature maps that is derived from empirical measurements of human retinal ganglion cells. We directly confirm this temporal difference experimentally by mutually exclusive stimulation of foveal and peripheral visual field regions. Last, we introduce a novel, data-driven method of recovering visual field information from neural data, highlighting and quantifying spatial, retinotopic information contained in temporally specific EEG recordings. Together, these results provide novel neural evidence for a temporal coarse-to-fine visual processing hierarchy in the processing of natural images that is directly linked to distinct spatial information sampling. Aligning the spatial sampling of humans and CNN encoding models not only improves predictions of neural responses but also demonstrates that EEG recordings contain a significant amount of temporally encoded retinotopic information. We make our large-scale EEG dataset including high-resolution natural scene images publicly available to enable future research into naturalistic visual processing.

Gold nanoparticles (AuNPs) are widely applied in nanomedicine, cellular and tissue biology, nanoscopy, photothermal therapy, and a range of diagnostic and clinical technologies. Among them, gold nanostars (AuNSs) have emerged as particularly promising due to their highly tunable optical and chemical properties. However, like other nanostructures, the stability of AuNSs remains a key challenge, especially within complex cellular microenvironments. Here, wide-field hyperspectral microscopy is evaluated for the real-time characterization of the morphology-dependent stability of AuNS formulations in immune-cell microenvironments. A computationally efficient image processing pipeline extracts statistical features from reflectance images, enabling the real-time analysis of hyperspectral data. UMAP-based visualization of spectral data revealed distinct, time- and formulation-dependent spectral shifts, with smaller seed volume formulations (larger overall diameter) for AuNSs exhibiting rapid destabilization and aggregation in THP-1 cells. In contrast, larger seed volume formulations (smaller overall diameter) for AuNS demonstrated enhanced colloidal stability and spectral uniformity. Compared to conventional ensemble measurements, hyperspectral reflectance measurements provided a rapid and resource-efficient approach that enabled macroscale imaging while retaining the spectral detail necessary to resolve AuNS transformations. Overall, the hyperspectral microscopy techniques presented here provide a label-free, high-throughput platform for evaluating AuNS stability and biocompatibility, with strong potential to guide the rational design of AuNSs for immunotherapeutic and diagnostic applications.