Large language models (LLMs) have demonstrated remarkable capabilities in understanding and generating human language from heterogeneous data sources. However, their integration with geospatial information remains limited, particularly in complex and dynamic environments where spatial reasoning must satisfy explicit physical, temporal, and operational constraints. This paper presents a constraint-aware Graph-based Retrieval-Augmented Generation (GraphRAG) approach for spatial reasoning in campus digital twin settings, designed to bridge unstructured text with geospatial topology. The framework integrates LLMs with a spatiotemporal knowledge graph that explicitly models campus facilities, events, accessible entrances, and real-time parking availability. Unlike traditional vector-based retrieval approaches, it leverages graph-structured retrieval to resolve critical constraints during the retrieval stage, thereby restricting downstream language generation to feasible and executable spatial configurations. This enables accurate responses to multi-criteria spatial queries, such as route planning and parking recommendations under multiple spatial and operational conditions. The framework is implemented as a constraint-aware spatial decision-support chatbot and integrated with an interactive visual analytics interface, allowing users to visually inspect the system’s reasoning process and outcomes. Experimental results demonstrate that the proposed GraphRAG approach significantly outperforms baseline retrieval-augmented generation methods in both accuracy and interpretability.

Organic upconversion devices (OUPCs) that convert near-infrared (NIR) photons into visible emission offer a promising platform for NIR visualization and detection, yet their practical deployment is hindered by high turn-on voltages and strong sensitivity to emitter composition. Here, we report a three-channel exciplex emitter architecture that enables low-voltage, emitter compositional-tolerant OUPCs. Benefiting from synergistic multichannel energy transfer and efficient reverse intersystem crossing within the exciplex system, the light-emitting unit exhibits a low turn-on voltage of 2.3 V and a maximum external quantum efficiency of 18%. Notably, nearly invariant emission performance is maintained over a wide D:A ratio range (DBT-SADF:PO-T2T from 2:8 to 8:2), with coefficients of variation of 0.11 for external quantum efficiency, indicating good processing tolerance. Based on this emitter, the resulting OUPC achieves an ultralow upconversion turn-on voltage of 1.6 V under 850 nm excitation, a maximum luminance on:off ratio exceeding 3.1 × 106, a detectivity of more than 1.5 × 1013 Jones, and visible emission (3 cd/m2) even at a weak NIR power density of 7 μW/cm2. Benefiting from the high-performance OUPC, we successfully achieved encrypted image decryption and bioimaging. These findings demonstrate that the proposed exciplex emitter strategy not only eliminates the reliance on heavy-metal emitters but also simultaneously enhances device performance and application potential, providing a scalable approach for NIR information visualization.

PocketMaster is a flexible and automated tool for the analysis, clustering, and interpretation of protein pockets, enabling the exploration of structural diversity in functional and interacting regions of proteins. The tool provides multiple strategies for defining pocket alignment regions, along with various alignment algorithms and clustering approaches, allowing analyses to be customized for different research objectives. In addition, it automatically documents results and provides informative visualizations and reports. With these capabilities, PocketMaster can be particularly valuable in the early stages of drug design, where accurate analysis and selection of protein structures are essential. Using TYK2 as a case study, PocketMaster demonstrates its ability to identify conformational differences between kinase and pseudokinase domains, as well as subtypes of structures within each domain, reflecting the influence of various ligands and protein states. The estrogen receptor alpha (ERα) ligand-binding pocket was also analyzed as an additional case study, showing the tool's performance in capturing conformational variations in helix 12 (H12) between active (agonist-bound) and inactive (antagonist-bound) states.The results confirm known structural features and illustrate the potential of the tool for systematic exploration of protein pockets, quantitative assessment of differences, and support of rational drug design. The PocketMaster source code, together with example input files and documentation, can be accessed at https://github.com/narek-abelyan/PocketMaster.

Fluorescent nucleobase analogs (FBAs) are valuable tools for studying nucleic acid structure and dynamics. However, their utility is often limited by substantial fluorescence quenching upon incorporation into oligonucleotides and variable brightness influenced by neighboring bases. In this study, we present a novel turn-on nucleoside, 3b, a thiazolyl-dU analog (hereinafter referred as TzdU), engineered to overcome these limitations and enable reliable DNA fluorescence imaging. Compared to its nearly nonfluorescent free form, TzdU shows approximately a 10-fold increase in brightness in single-stranded DNA (ssDNA) and up to a 50-fold enhancement in double-stranded DNA (dsDNA). Importantly, it maintains relatively stable brightness regardless of surrounding bases by evading common quenching pathways, including solvent-induced collisional quenching and excited-state proton transfer (ESPT). The triphosphate derivative of TzdU is efficiently utilized by various DNA polymerases, including Deep Vent and KOD XL, facilitating real-time, intensity-based monitoring of critical enzymatic processes such as PCR and primer extension without external labels. Furthermore, TzdU can illuminate DNA in a gradient manner, enabling the visualization and encryption of information. As the first FBA to achieve universal turn-on characteristics, sequence insensitivity, and compatibility with enzymatic reactions, TzdU serves as a novel tool for investigating nucleic acid dynamics and advancing fluorescence-based methodologies.

This paper presents ALFA DT, a Digital Twin (DT) designed for Assembly Line Fitness Analytics (ALFA) to support descriptive, predictive, and prescriptive analytics for assembly line performance. The paper details the digitalisation of the assembly line process, focussing on generating digital models of operator walk patterns and creating task precedence constraints. Two algorithms are proposed, one called Walk Pattern (WP) for the digitalisation of operator walk patterns in the line and another called BiDirectional Precedence Miner (BDPM) algorithm for the precedence graph generation considering quick and frequent product model evolution, which are two of main challenges in current industries. The implementation of ALFA DT is based on the UNITY game engine, leveraged for its visualisation and animation capabilities, multi-platform deployment, and collaborative environment. The proposed DT provides an efficient tool for analyzing, forecasting, and optimising assembly line performance. ALFA DT integrates real-time data and learns from aggregated system behaviour to support predictive insights and proactive decision-making. By serving as a single source of relevant data with an interactive visual interface, it facilitates advanced analytics applications, allowing users to easily validate and modify proposed solutions. ALFA DT has been applied on our real automotive assembly setting and the findings indicate that it improves operational decision-making by improving data accessibility and visualisation quality. The paper discusses several use cases of ALFA DT, including operator work pattern analysis, workcells' layout optimisation, simultaneous assembly line balancing and layout adjustment, and predicting operator overload.

BackgroundAn increasing amount of digital health data are being collected across rehabilitation settings, but their integration into routine clinical practice remains limited, despite its potential to motivate patients or inform clinical decision-making. Specifically, effective visualization and communication of assessment outcomes to both patients and health care practitioners (HCPs) represent a key gap in the neurorehabilitation practice.ObjectiveThis study describes the development and evaluation of RehaLink (author ND, ETH Zürich), a proof-of-concept mobile app that delivers structured, interpretable feedback from conventional and technology-based assessments to neurorehabilitation patients and HCPs.MethodsThe app was developed through a 3-step iterative co-design process involving 17 inpatients with multiple sclerosis and 15 HCPs from a single rehabilitation center. The app integrates a full battery of conventional assessments routinely conducted at the clinic, as well as digital health metrics from the Virtual Peg Insertion Test, a validated technology-based assessment of upper limb function, as a proof of concept for integrating technology-based assessment data into clinical workflows. Three structured feedback sessions were conducted, in which participants evaluated feedback types, visualization formats, and app usability using Likert-scale ratings, preference rankings, open-ended questions, and the System Usability Scale. Data were analyzed using descriptive statistics and directed content analysis.ResultsAcross all 3 sessions, progress bars and color-coded indicators were consistently preferred over text-heavy or abstract formats by both patients and HCPs. A persistent set of competing demands was observed, with participants requesting both visual simplicity and access to absolute values and normative comparisons. HCPs tended to underestimate patients’ preference for informative visualizations. The perceived value of structured feedback increased over the course of the study; patients’ median ratings rose from 4.0 to 5.0 and HCPs’ from 4.0 to 4.5 on a 5-point Likert scale. The resulting mobile app prototype demonstrated high usability, with patients achieving a mean System Usability Scale score of 93.6 (mean 6.4; best imaginable) and HCPs 80.9 (SD 8.1; good), according to established benchmarks.ConclusionsThese findings demonstrate the feasibility and value of a co-designed digital feedback tool for neurorehabilitation. By combining conventional and technology-based assessment outcomes in an accessible, user-centered format, the app has the potential to enhance patient engagement, support clinical decision-making, and advance the implementation of value-based, personalized care.

Understanding complex health information, such as kidney function values (eg, creatinine), is important for youth kidney transplant recipients and caregivers to effectively engage and participate in their care. Information visualizations, such as visual analogies, highlight the similarities between 2 different ideas through visual means and can support understanding of abstract data to facilitate self-management. The study was motivated by the persistent challenge that youth and caregivers face in interpreting complex clinical data, which often remains unactionable and disconnected from their practical information needs. This study aims to design and evaluate a novel visualization to support the information needs of youth kidney transplant recipients and caregivers, with the goal of increasing actionable and accessible knowledge for self-management. We conducted 2 studies with youth kidney transplant recipients and their caregivers: study 1 identifies visualization elements that support patient and caregiver kidney function information needs, and study 2 evaluates a novel patient-facing kidney function visualization informed by results from study 1. In study 1, participants drew representations of their kidney function over time. These findings led to the development of the novel kidney bean visual. In study 2, participants interacted with the personalized (incorporating electronic health record data) kidney bean visual component for 1 week, then completed a feedback session exploring how the novel visual facilitated retrospective reflection of their health status and the specific behaviors influencing their renal function. We conducted qualitative content analysis on all interviews. All sessions and interviews were recorded, transcribed, and analyzed using content analysis (study 1), thematic analysis, and affinity diagramming (study 2). In study 1 (n=30), youth and caregivers consistently described the difficulty of understanding their creatinine results and related medical terminology, with the majority using emotions to represent their kidney function. Therefore, our kidney bean visual, a visual analogy, incorporated emotions and colors correlating with their creatinine levels and overall kidney function. In study 2 (n=23), we found that for many participants, the kidney bean visual (1) mirrored and validated their emotions, prompting reflections on their kidney function contextualized by their lived experiences; (2) served as a progress check; and (3) acted as a motivator for healthy kidney care habits. Kidney function laboratory values, although seemingly straightforward in their intended purpose and use, are more complex from the patient and family perspective, as they also closely align with youth kidney transplant recipients' and their families' emotions. This study demonstrates the value of designing visualizations that capture these complexities, which can create opportunities for increased understanding of health information and self-reflection to support self-management.

Computed tomography (CT) imaging is essential for lung cancer (LCa) management, while offering detailed visualization and valuable information for Artificial Intelligence-guided prognosis. Meanwhile, supervised learning (SL) models require extensive labeled data, limiting their real-world utility where annotations are scarce. We analyzed CT scans from 977 patients across 12 public/private datasets, extracting 1,218 radiomics features using Laplacian of Gaussian and wavelet filters via standardized PyRadiomics. Dimensionality was reduced using 56 feature selection and attribute extraction algorithms, and 27 classifiers were benchmarked. Semi-supervised learning (SSL) framework with pseudo-labeling utilized 478 unlabeled and 499 labeled cases. Model sensitivity was assessed in three scenarios: varying labeled data in SL, increasing unlabeled data in SSL, and jointly scaling both from 10% to 100%. SHapley Additive exPlanations analysis (SHAP)analysis explained and interpreted the top model predictions. Five-fold cross-validation and external testing (in two cohorts) were performed. SSL outperformed SL across all metrics, improving overall survival prediction by up to 17%. The top SSL model (Feature Importance by Random Forest+XGBoost) achieved 0.90±0.01 accuracy in cross-validation and 0.88±0.01 externally. SHAP revealed enhanced feature discriminability in SSL and SL, particularly for class 1 (survival>4 years), and helped explain model decisions. SSL maintained strong performance even with only 10% labeled data. Both SSL-based scenarios demonstrated more stable performance compared to SL, with lower variance across external testing, emphasizing SSL's robustness and cost-efficiency. We propose an interpretable, cost-effective SSL framework for CT-based LCa survival prediction. It improves performance, generalizability, and clinical readiness via SHAP explainability and unlabeled data utilization.

<p>This article examines the impact of visual techniques, such as bold, underline, and italics, on the effectiveness of processing legal document content. The study empirically tests Richard E. Mayer’s signaling principle, which posits that highlighting key information supports comprehension and recall. An experiment conducted with non-law students revealed that the effectiveness of typographic cues is not uniform and varies depending on the syntactic complexity of the legal content. Participants performed better with simpler linguistic structures, regardless of the type of visual emphasis, while increased complexity diminished the effectiveness of all signaling methods. These findings suggest that, under the conditions of this study, the effectiveness of typographic signaling is limited and strongly dependent on linguistic complexity. While the results do not invalidate the signaling principle, they indicate that its effects may be constrained in contexts involving cognitively demanding legal texts.</p>

In today's highly competitive business environment, customer retention is a critical driver of long-term profitability and sustainable growth. Identifying at-risk customers before they disengage allows businesses to implement proactive, targeted retention strategies rather than relying on reactive measures. This project presents the development of a Real-Time Churn Prediction Tool and an accompanying interactive dashboard designed to empower customer success teams with actionable insights. Leveraging machine learning techniques, the underlying predictive model analyzes historical customer data—including demographic profiles, engagement metrics, and service usage patterns—to calculate the probability of churn for individual users. The predictive insights are surfaced through a dynamic, Streamlitbased web application that provides real-time visualization of key performance indicators (KPIs) and underlying risk factors. The dashboard features an intuitive user interface that allows stakeholders to seamlessly filter data, monitor overall churn metrics, and drill down into specific high-risk customer segments. By synthesizing predictive analytics with an accessible visual interface, this system facilitates data-driven decision-making. Ultimately, the tool enables organizations to optimize resource allocation, enhance

People reliably associate the meanings of both abstract and concrete words with colors distributed over color space, a phenomenon that influences aspects of visual cognition ranging from object recognition to interpreting information visualizations. Prior research has hypothesized that color-concept associations arise from the cross-modal statistical structure of experience, but it remains unclear whether natural environments contain such structure or whether learning systems can discover it without strong prior constraints. To address these questions, we investigated whether GPT-4, a multimodal large language model, can estimate color-concept association ratings that approximate those made by people. We tested 71 colors spanning perceptual color space and a variety of concepts varying in abstractness. GPT-4 ratings correlated strongly with human ratings across a range of prompting strategies, outperforming prior state-of-the-art methods for automatically estimating color-concept associations from images. In an empirical study assessing people's ability to interpret the meanings of colors in information visualizations, palettes generated from GPT-4's rating data were not only interpretable but, in some cases, more effective than those based on human ratings. Taken together, our results suggest that high-order covariance between language and perception, present in web-scale data, provide sufficient information to learn color-concept associations without initial constraints, and that machine-derived associations can support the optimization of information visualizations for visual communication.

A systematic review of surgeon-anesthesiologist relationship in the operating room.

The time dimension matters: Improving mode of action classification with live-cell imaging

ABSTRACT In smart spaces, multiple connected objects and sensors act in a coordinated manner through automations that are deployed into the environment. A common approach to empowering users to configure the desired automated behaviours is using trigger-action programming (TAP). However, it can be difficult for users to untangle the behaviour of the created rules. In this paper, we propose CASPER, a novel tool to manage smart environments, support users in controlling their objects and devices, and create and modify automations. It provides support in detecting and understanding possible problems between TAP rules (such as conflicts and unexpected direct or indirect rule activations), and between such automations and users’ goals (e.g. energy, wellbeing, health). It also provides suggestions for potential improvements to current automations. The feedback from the evaluation of an initial prototype and the analysis of the state of the art guided the development of the new platform, which integrates conversational interaction and a visual interface, thus providing a comprehensive IoT automation management environment. CASPER was then tested in a follow-up study involving 16 participants. The revised prototype received good usability ratings, unveiled interaction patterns across the two interaction modalities, and informed design implications for further developments.

Three-dimensional (3D) geological numerical models and control information models play important roles in numerical analysis, engineering information organization, and visual representation, and are therefore essential in intelligent engineering and geological hazard assessment. However, during the mapping of numerical results from numerical models to control information models, differences in data structures, topological organization, attribute representation rules, and storage formats often lead to "broken chain" problems, such as discontinuous attribute information, missing structural relationships, and abnormal geometric representation, thereby affecting the accurate expression and effective application of numerical results. To address this issue, this study proposes a rule-driven broken-chain correction method for mapping between 3D geological numerical models and control information models. Focusing on the one-way transfer of numerical simulation software results to a front-end control information model, the proposed method identifies data variations at both the attribute and structural levels during the mapping process, and combines rule-based correction with prior-mesh topology reconstruction to achieve reliable representation of numerical results in the target environment. The reconstructed nodes, patches, and attribute information are further organized into structured data that support loading, visualization, and information management. Using a pile-foundation engineering case in Southwest China, the proposed method was validated based on an numerical model and a front-end control information model implemented with JSON and Three.js/WebGL. The results show that the method can effectively restore model topology, suppress local topological disorder and self-intersection, and preserve the key numerical characteristics of nodal attributes during cross-platform mapping. Further reliability analysis indicates that, under attribute datasets of different sizes, the proposed method exhibits good topology correction performance and stable attribute mapping while satisfying the requirements of visualization. This study provides a feasible technical pathway for the cross-platform transfer of 3D geological numerical results to control information models, and also offers methodological support for the organization, visualization, and collaborative application of geotechnical engineering information in multi-software environments.

The traditional monitoring of forest musk deer behavior primarily relies on direct human observation or the post hoc playback analysis of ordinary surveillance videos. This approach is not only time-consuming and labor-intensive but also highly subjective, easily leading to missing or misjudged critical behavioral information. Moreover, it is difficult to achieve real-time monitoring and anomaly warning. These limitations severely constrain the efficiency of the large-scale artificial breeding of forest musk deer and the effective advancement of wild population conservation. Thus, this study proposes a forest musk deer behavioral recognition method based on an improved YOLOv8-Pose. A forest musk deer behavior image dataset covering four typical behaviors was constructed, and 18 keypoints were systematically annotated. This study designs a Dilated Spatial Pyramid Pooling-Fast (DILATED-SPPF) module and a Multi-scale Depthwise Separable Context Mixer (MDSC-Mixer) module, and integrates them into YOLOv8-Pose. Experimental results show that the improved model outperforms the original YOLOv8-Pose and comparison models such as YOLOv11/v12-Pose on key metrics of object detection (Box-mAP50 0.929, Box-mAP50-95 0.814) and pose estimation (Pose-mAP50 0.879, Pose-mAP50-95 0.565). This study further develops a visual interactive interface that intuitively presents detection results and skeleton structures. This work provides a high-precision, low-cost automated behavior analysis tool for the artificial breeding and wild conservation of forest musk deer with significant application value for enhancing the intelligence level of endangered species protection.

Diagnosis of digestive tract (DT) diseases is limited by invasive wired endoscopy and capsule endoscopy (CE) challenges, including exogenous fluorescence risks, poor WLI contrast, and limited information acquisition of narrow-band modes. This study presents label-free imaging CE to noninvasively detect DT vessels, improving diagnostic accuracy and patient safety. Multispectral imaging CE (MICE) allows conventional white-light examination with a single light source, as well as depth-specific imaging of superficial capillaries (420 nm), shallow-seated branches (540 nm), and deep, thick vessels (590 nm), improving visualization and depth differentiation of mucosal vascular morphology. Furthermore, through synergistic multispectral operation, MICE facilitates blue-light imaging (BLI: white + 420 nm), narrow-band imaging (NBI: 420 nm + 540 nm), and dual-red imaging (DRI: 540 nm + 590 nm + 630 nm), capturing multilayered, morphologically rich vascular information for the comprehensive assessment of DT diseases. Proof-of-concept evaluations under 420, 540, and 590 nm illuminations successfully captured depth-varying vascular information on porcine DT mucosa. The vessel-to-background contrast consistently improved over WLI, with gains of 2.90, 1.94, and 1.21. Furthermore, fusion visualization of multilayered vascular information was accomplished via BLI, NBI, and DRI. In addition, an integrated platform, featuring wireless power transfer (WPT) and magnetic control (MC), ensures a stable energy supply for MICE while enabling precise control over its locomotion and orientation. To our knowledge, this represents the first label-free MICE for noninvasive detection of DT vasculature, holding significant clinical implications for diagnosing DT diseases.

Purpose:We present a large-scale (N= 120) comparative study of gel-based and dry electroencephalography (EEG) systems for cognitive load analysis in tasks involving information visualization (InfoVis) stimuli. Although dry systems are increasingly adopted owing to their portability and fast setup, their sensitivity to cognitive-related measurements (as compared to gel-based systems) remains debated. This limits the understanding of whether dry systems provide sufficient sensitivity for cognitive load assessment under controlled task conditions.Methods:We analyzed a diverse set of signal quality metrics, such as signal-to-noise ratio and channel retention, combined with spectral features across frequency bands to evaluate the ability for each device to capture workload-related neural markers during InfoVis tasks.Results:Although the gel-based device showed consistently better quality results than the dry one, the effect sizes suggest a small practical significance of the differences between systems. These results demonstrate that dry systems can provide adequate physiological sensitivity for cognitive load assessments.Conclusion:Our findings highlight the trade-off between usability (setup, calibration, etc) and data fidelity, providing practical guidance for choosing EEG systems for cognitive workload monitoring and applied neuroengineering research. Overall, the results suggest that dry systems can support coarse-grained cognitive load assessment, while gel-based systems remain advantageous when greater sensitivity is required.

CYP450 (cytochrome P450)-mediated drug metabolism is a critical determinant of pharmacokinetics and clinical safety, making comprehensive metabolic profiling essential for rational drug discovery. Here, we present DeepCYP (https://deepcyp.scbdd.com), a freely accessible deep-learning web server for end-to-end CYP450 metabolic profiling. Trained on an expanded dataset and a mechanism-based reaction rule library, DeepCYP uses a multi-task graph neural network (GNN) combined with multi-scale descriptors. Operating directly on 2D molecular graphs, this architecture bridges the entire "pathway-site-product" continuum across nine major CYP isoforms (CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2E1, and CYP3A4) within a unified pipeline. Benchmarking demonstrates that DeepCYP outperforms established tools, including FAME3, SMARTCyp, and BioTransformer 3.0, improving Top-1 and Top-2 ranking metrics by over 10%. Furthermore, the server supports high-throughput batch processing, capable of evaluating ~280 molecules per minute. DeepCYP also enhances interpretability through an interactive visualization interface, featuring susceptibility radar charts and dynamic transformation tables. By translating abstract predictions into biological insights, DeepCYP provides a practical tool to accelerate lead optimization and mitigate toxicity risks.

Objective: To develop a computational tool for the integration and visualization of agronomic and cartographic data in GeoJSON format, applied to agricultural irrigation management. Design/methodology/approach: A descriptive and applied approach was adopted through the design of a Python-based system that links spatial information in shapefile format with tabular agronomic data derived from spreadsheets. The process incorporates identifier normalization, tabular-to-spatial joining, and automated export in compliance with the RFC 7946 standard. The tool was implemented using real data from the Batequis Irrigation Module of Irrigation District 075, located in northern Sinaloa, Mexico. Results: The system generated an interoperable GeoJSON file that preserved the geometric integrity of agricultural plots and the consistency of agronomic attributes. Geospatial visualization enabled the interpretation of water stress distribution at the plot level. The usability evaluation yielded an average score of 82.5 on the System Usability Scale (SUS), while the functional evaluation, based on ISO/IEC 25010, showed a compliance level of 94%. Limitations on study/implications: The tool was applied in a single irrigation module; therefore, its performance is associated with the operational and agronomic conditions of that specific context. Its application in other irrigation modules and districts is recommended in order to broaden the scope of the results. Findings/conclusions: The developed tool provides a robust technical foundation for strengthening the digitalization of irrigation services through interoperable geospatial products. The proposed solution is both viable and pertinent for enhancing the integration, visualization, and analysis of agricultural information, thereby contributing to more informed decision-making in irrigation management.