This paper presents an integrated assistive approach that combines multimodal vision-language models with advanced computer vision techniques to support visually impaired individuals. The proposed system utilizes object detection via a custom-trained YOLOv8 model on an expanded dataset, along with MiDaS for monocular depth estimation. This enables the extraction of visual cues about surrounding objects and their relative distances, allowing near real-time hazard recognition and contextual descriptions via voice feedback. The system is deployed on a cloud/server infrastructure using a lightweight prototype (Google Colab + Ngrok), which introduces an average latency of 15–17 seconds per response. A dedicated Vietnamese dataset of annotated images, warnings, and context-specific descriptions was developed. To evaluate semantic alignment between model-generated and human-written descriptions, cosine similarity (using SBERT embeddings) achieved approximately 95%, far above the 0.5 threshold. Natural Language Inference (NLI) techniques were used to assess logical consistency. For overall descriptions, 72.5% were labeled neutral, 14.5% entailment, and 13.0% contradiction. For hazard warnings, 73.5% were entailment, 25.5% contradiction, and only 1.0% neutral. These results demonstrate that the model produces reliable hazard descriptions while general summaries may omit minor contextual details. In conclusion, the integration of vision-language models with object detection and depth estimation offers a scalable and effective assistive solution for the visually impaired. The system achieves high semantic fidelity in descriptive tasks and robust hazard communication, proving its potential for real-world deployment in accessibility technologies.

Environmental quality indicators are essential tools for translating complex technical data into accessible information, facilitating communication between academia, the public, and policymakers. Among these indicators, water quality plays a crucial role in public health and environmental preservation, particularly in ecologically sensitive regions such as the Amazon. This study presents a comparative analysis of water monitoring methods in the city of Parintins, Amazonas, Brazil, contrasting an automated data collection device with embedded sensors and traditional manual sampling using pre-calibrated commercial probes. Prior to this initiative, monitoring in the region relied exclusively on manual measurements performed by local researchers, limiting both the frequency and scope of analysis. The proposed prototype was deployed on a floating platform at the Port of Parintins and operated using LoRa-based data transmission to ensure functionality in areas with limited connectivity. The monitored parameters included pH, turbidity, electrical conductivity, water temperature, and dissolved oxygen. Data collected automatically were compared with those obtained from manual probes to assess the precision, stability, and feasibility of the automated system. Our results revealed a strong correlation between both methods, with the automated approach demonstrating superior consistency and frequency of measurements. The findings support the use of the proposed system as an efficient, low-cost, and reliable alternative for continuous water quality monitoring in remote regions of the Amazon, contributing to sustainable environmental management and informed decision-making.

The ultra-sensitive quantification of cytokines is a critical demand in biomedical diagnostics, particularly in monitoring immune dysregulation, chronic inflammation, and cancer-related signaling. This short overview presents two low-cost, simple, ultra-fast response, and small-size plasmonic point-of-care tests (PoCTs), both specifically developed for the attomolar detection of substances of interest, such as the key interleukins (e.g, IL-17A, IL-1β, and IL-18), demonstrating outstanding analytical performance without the need for amplification protocols.

The first PoCT developed by Cennamo’s Lab is a microcuvette-based plastic optical fiber (POF) device. The device exploits the multimodal POFs characteristic to change via analyte-receptor binding, the plasmonic phenomena in an SPR D-shaped POF probe arranged in series. The system operates without surface functionalization; indeed, the bioreceptor–analyte interaction occurs freely in solution within modified POFs (the microcuvette)¹. The second Cennamo’s PoCT exploits a nanoplasmonic biosensor chip based on pollen nanostructures covered by gold nanofilms and functionalized with specific antibodies. This architecture enables the excitation of hybrid plasmonic modes, yielding unprecedented sensitivity for cytokine detection via antibodies down to the sub-attomolar range².

Both sensing approaches exhibit rapid response times (approximately 10 minutes), high molecular specificity, and excellent reproducibility, making them highly attractive for the real-time monitoring of cytokines in diluted biological fluids or other substances of interest at the single-molecule level via PoCTs.

References

¹ N. Cennamo, F. Arcadio, C. Marzano, R. Pitruzzella, M. Seggio, M. Pesavento, S. Toldo, A. Abbate, L. Zeni, Sensors 2025, 25, 930.

² C. Marzano, R. Pitruzzella, F. Arcadio, F. Passeggio, M. Seggio, L. Pasquardini, N. Cennamo Biosensors 2025, 15, 161.

Background: Currently, in medical analysis, the mortality rate is enhanced due to cardiovascular diseases. The WHO reports that approximately 17.9 million people are afflicted with heart diseases, which leads to death across the globe. Cardiovascular diseases are illnesses that affect the heart and blood vessels, including cerebrovascular disease, coronary heart disease, rheumatic heart disease, etc. Objective: In this article, our objective is to predict early heart disease using deep learning. We designed a framework that addresses the critical gaps in heart disease prediction, as well as ensuring privacy, clinical utility, and equity. Materials/methods: We used several classification models, such as LogR, RF, XGBoost, 1D-CNN (Centralized), LSTM (Centralized), FedAvg (EHR Only), FedHeart-MM and FedHeart-MM + DP (ε=2) to predict early heart disease. We proposed a novel framework called FedHeart-MM to achieve this task. The FedHeart-MM architecture was designed to facilitate distributed training among several client nodes while maintaining the confidentiality of raw patient data. The federated configuration enables each client (e.g., hospitals or data centres) to conduct local training on their data while transmitting just model parameters to a central server. Result: This study's proposed model, FedHeart-MM achieved 95% AUC-ROC, sensitivity of 92%, specificity of 93%, and an F1-score of 82%. In comparison to another model, our proposed model performs well and identifies early heart disease. We also observed that there is a 14% higher AUC vs. our traditional approaches. Federated models exhibited enhanced privacy-preserving capabilities.

Phenomena of damaged ecosystems and environments have caught the widespread attention of scientists and engineers. Earthquake events are good examples, severely damaging slopes and houses. A study is conducted on slope stability at earthquake limit state so that slopes and relevant houses can be safeguarded. The novel analytical and numerical methods are applied to the study on slope stability at earthquake limit state. The analytical method consults moment equilibrium equations of slopes and a limit equilibrium theory. The numerical method uses the coupled finite elements and strength reduction approach at the earthquake limit state. In addition, earthquake effect for the analytical and numerical methods is considered as a horizontal static force through the pseudo-static method. The validation of the analytical and numerical methods is verified through a comparison between their calculation results. In engineering examples at the earthquake limit state, factors of safety of slopes are shown versus various strength factors of slopes and earthquake intensities. Factors of safety of slopes increase with the increase in strength factors of slopes, including equivalent friction angles and cohesive strength. However, factors of safety of slopes decrease with the increase in earthquake intensities. They imply slopes may be more safe with strong strength factors and low earthquake intensities. These would provide references for the protection of ecosystems and environments.

This study addresses a critical gap in seismic design by quantifying how plan asymmetry and multiple earthquake sequences interact to affect the nonlinear response of reinforced concrete (RC) frames. While earthquake-resistant design provisions have evolved, most current codes are based on single-event assumptions and simplified symmetry considerations, overlooking the cumulative effects of repeated ground motions observed in recent international studies.

In this research, symmetrical and asymmetrical low-rise RC frames are analyzed through nonlinear dynamic simulations, considering both single-event and multiple-event ground excitations for comparison. The analyses incorporate three-dimensional ground motions in horizontal and vertical directions, while explicitly modeling the nonlinear inelastic behavior of RC sections under severe seismic demands.

A dimensionless asymmetry ratio, defined by the relative distribution of stiffness and mass, is proposed to systematically evaluate torsional sensitivity in plan-irregular structures. Structural responses are further expressed through normalized indices, which enable consistent comparisons across different frame geometries and seismic scenarios.

The results show that increasing plan asymmetry amplifies inter-story drift, torsional rotations, and plastic hinge concentrations, particularly under successive earthquake sequences. These findings indicate that existing design provisions may underestimate the vulnerability of irregular RC buildings. This work is among the first to integrate plan asymmetry and multi-event seismic loading into a unified evaluation framework, offering a novel tool for refining earthquake-resistant design standards.

This work introduces improved estimators for fourth-order cumulants in bi-additive models involving both fixed and random effects. Cumulants are powerful statistical measures that extend beyond traditional moments, offering more nuanced descriptions of probability distributions, including characteristics such as skewness and kurtosis. By utilizing the cumulant-generating function and least-squares methods, we develop advanced estimators that enable accurate inference from independent and identically distributed (i.i.d.) data. The proposed bi-additive models, which incorporate deterministic fixed components and independent random terms, facilitate precise estimation of fourth-order cumulants while accounting for the variability commonly observed in real-world data. These models also enable the systematic analysis of distributions with parameters related to location, dispersion, and shape, thus shedding light on their underlying structure. We demonstrate that the proposed methods offer both theoretical robustness and practical utility across diverse applied contexts. We illustrate the applications of this methodology in two domains: anomaly detection in sensor networks, where higher-order cumulants help identify deviations from expected patterns, and variability analysis in materials science, where they capture subtle differences in material properties. Our results emphasize the significance of higher-order cumulants in revealing complex features in data that are often obscured in mean-variance-based analyses. This contribution emphasizes both the theoretical significance and practical benefits of adopting advanced cumulant estimators in bi-additive models, equipping researchers in the applied sciences with novel tools for exploring and understanding the complexity of empirical distributions.

The controlled release of nitric oxide (NO) within tumor microenvironments is an attractive strategy in anticancer therapy, as NO displays dose-dependent cytotoxic effects. Light-activated NO donors provide unique spatiotemporal precision and minimal invasiveness, making them highly suitable for biomedical applications.

Here, we describe NCDs-1, a water-soluble nanoconjugate obtained by coupling a two-step NO photodonor with blue-emitting nitrogen-doped carbon nanodots (NCDs). The resulting hybrid nanostructure (~10 nm) exhibits a new absorption band that is absent in the individual components, indicating a strong ground-state electronic interaction. Upon blue-light irradiation, NCDs-1 achieves nearly tenfold higher NO release than the free photodonor, which is a result attributed to the photoinduced electron transfer between NCDs and the NO-releasing unit. Importantly, the quenched fluorescence of NCDs is restored during the second NO release step, thus providing a real-time optical signal for monitoring the process. In addition, NCDs-1 shows remarkable photothermal conversion efficiency, supporting its potential for multimodal therapy.

To shift the light responsiveness toward more biocompatible wavelengths, we developed NCDs-2 by modifying the solvent conditions during the synthesis of NCDs while retaining citric acid and urea as precursors. This adjustment generated NCDs with absorption shifted into the green region. When conjugated with the same NO donor, NCDs-2 maintained efficient NO release upon green-light irradiation, a wavelength with improved tissue penetration and compatibility.

Preliminary in vitro experiments on cancer cell lines confirmed the therapeutic promise of both nanoconjugates. Collectively, NCDs-1 and NCDs-2 represent versatile multifunctional platforms for light-controlled NO delivery and combined photothermal therapy, with tunable optical properties adaptable to cancer diseases.

Quantum computing is poised to reshape the landscape of modern computation by offering exponential advantages in domains such as cryptography, optimization, and intelligent data processing. To realize the full potential of quantum systems, especially in fault-tolerant and Noisy Intermediate-Scale Quantum (NISQ) environments, it is essential to design quantum circuits that are both resource-efficient and resilient to errors. This paper presents a novel design of a Binary-Coded Decimal (BCD) to Excess-3 code converter circuit using only the Clifford+T gate set, a universal gate library widely supported on current quantum hardware. Our approach replaces conventional 4-bit reversible adder-based implementations with an optimized logic design based on Clifford+T-decomposed Peres gates. The use of Temporary Logical-AND gates and CNOT operations significantly reduces the T-count, circuit depth, and quantum cost—key metrics for fault-tolerant quantum computation. The circuit is simulated using IBM Qiskit, and functional verification across all valid BCD inputs confirms the correctness of the design. Beyond its core arithmetic function, the converter has broader implications in AI-oriented quantum architectures, where low-latency and low-error arithmetic operations are vital. This work contributes a scalable, fault-tolerant, and hardware-compatible building block for quantum logic, paving the way for practical and resource-aware quantum circuits in future AI-enhanced quantum systems.

We report in this study novel topology of terahertz imaging system, including a single-radar multi-output guided system. We propose the integration of a frequency modulated continuous wave (FMCW) radar at 120 GHz with multiple plastic waveguides allowing to do sensing in reflection mode with multiple ouputs. The FMCW radar, from Indie Semiconductor, which has an antenna on package, is directly coupled to nine HIPS waveguides which are placed front of the package. These different waveguides, made with additive manufacturing HIPS, then move away in three dimensions, each have a different distance and go towards a line on which they are all aligned, to take a measurement in reflectometry. The radar then mesure the reflected signal, and after processing, we extract the time of flight. The difference of each path allow to determine the peak coming from each pixeL This simple and direct coupling allows a compact setup without requiring expensive metal waveguide and couplers. This speed up by nice the imaging measurement which can be useful for FMCW radar imaging application like non-destructive testing, particularly in industrial process and inline control. We propose to present both tinite difference time tomain simulation with CST Mcrowave Studio software, experimental integration and results ans discussion.