IEEE Latin America Transactions

A Decentralized Learning Architecture for Medical Prescription Anomaly Detection via Hybrid Federated-Swarm Learning

Ravelly Zanatta — 2026-07-14

The increasing use of electronic medical records (EMRs) has improved efficiency, accuracy, and accessibility of patient data. However, conventional centralized architectures suffer from single points of failure and data privacy issues. To address these challenges, this study proposes a decentralized machine learning architecture that combines concepts from Federated Learning (FL) and Swarm Learning (SL) for anomaly detection in medical prescriptions. The proposed architecture leverages blockchain and the InterPlanetary File System (IPFS) to enable secure model sharing and decentralized storage, thereby reducing communication complexity and establishing a transparent, decentralized parameter repository. Experimental evaluations were conducted using logistic regression (LR), a multi-layer perceptron (MLP), and a decision tree (DT) model. Compared with the FL baseline, the proposed system achieved superior efficiency, lower resource consumption, and improved latency, along with smaller block sizes. It, however, exhibited slightly lower transaction throughput and longer training rounds, reflecting the added complexity of decentralization. In predictive performance on the anomaly classification task, DT achieved the highest precision and recall under the evaluated dataset (F1-score=0.9912), followed by MLP (0.5504) and LR (0.2442). The decentralized training approach led to negligible performance loss relative to centralized models, less than 4% for LR and below 1% for both MLP and DT. Overall, the proposed system demonstrates a robust and efficient alternative for decentralized learning in healthcare applications, maintaining strong predictive performance while enhancing architectural transparency.

Clinical Practice Guidelines as a JSON Service: A Proposed Architecture for Decision Support in Major Burning Management

Jose de Jesus Alvarez Ramirez — 2026-07-14

Clinical Practice Guidelines (CPGs) encode evidence-based clinical knowledge but are primarily distributed as unstructured PDF documents, making them inaccessible to automated clinical decision support (CDS) systems. This paper proposes a service-oriented architecture that formalizes the IMSS Clinical Practice Guideline for Major Burn Management (IMSS-375) as a versioned REST/JSON microservice. Seven clinical decision endpoints are defined, each encapsulating a specific GPC recommendation: burn classification, initial assessment, fluid resuscitation, pain management, infection prevention, nutritional support, and transfer criteria, following HL7 FHIR R4 interoperability standards. A mapping between GPC clinical rules (including the Parkland formula, Benaim scale, Curreri formula, and Baux prognostic index) and structured JSON request/response schemas is presented and evaluated against related formalization approaches and verified through structured schema-level invocations against a representative clinical scenario, including a detailed comparison of Mexico IMSS-375 standard properties against HL7 FHIR and OpenEHR. A deployment architecture is described covering hospital-level integration, a centralized service layer, and a non-relational persistence tier based on document-oriented storage for unstructured clinical data.

DMODBUS: Blockchain for Network Identity Validation in Modbus Connections

Paulo Henrique Mariano — 2026-07-14

The Fourth Industrial Revolution, driven by the convergence of Information Technology (IT) and Operational Technology (OT), has accelerated industrial innovation while exposing legacy systems to modern cyber threats. Widely adopted protocols, such as Modbus/TCP, lack native security mecha nisms and remain inherently vulnerable to exploitation. This study proposes and validates DModbus, an adaptive security layer that enhances Modbus communications by integrating blockchain technology. The architecture introduces a non-invasive gateway that utilizes a permissioned Proof of Authority (PoA) blockchain as a distributed and immutable ledger for device identities, enabling robust authentication during communication. The proposal was validated through a Proof of Concept (TRL 3) simulating a Man-in-the-Middle (MitM) attack via Address Resolution Protocol (ARP) spoofing in a Modbus TCP/IP envi ronment. The experimental results demonstrate that DModbus detects identity spoofing attempts in real time. Performance analysis quantifies the inherent trade-off: an average latency overhead of 91.3% relative to the unprotected baseline. This overhead suggests suitability for non-critical real-time applica tions, such as supervisory control and data acquisition (SCADA) systems, rather than high-speed actuators, given the significant gains in security and resilience—particularly when contrasted with the severe and unpredictable degradation observed during active attacks. The main contributions are: (i) a model for securing Industry of Things (IIoT) devices using blockchain; (ii) a quantitative assessment of Proof-of-Authority (PoA)-based overhead in OT networks; and (iii) validation of decentralized identity management as a viable security layer for industrial communications.

Use of Radio Frequency Identifiers in Autonomous Mobility Systems

João Mota Neto — 2026-07-14

Autonomous mobility systems rely heavily on perception technologies to interpret road infrastructure and make driving decisions without human intervention. Among these technologies, camera-based traffic sign recognition systems are widely employed; however, their performance is significantly degraded under adverse conditions such as sign deterioration, partial occlusion, poor lighting, and vandalism. In countries with limited infrastructure maintenance, these conditions represent a critical safety risk. This work proposes and experimentally evaluates the use of Radio Frequency Identification (RFID) technology as a redundant perception layer for traffic sign identification in autonomous mobility systems. Passive UHF RFID tags were integrated into traffic signs and detected using a vehicle-mounted reader system. The experimental evaluation was conducted in two stages: controlled bench tests and field tests in a real campus environment. Bench tests resulted in a mean recognition distance of 10.35 m with a 95 \% confidence interval of ±0.40 m, while field tests yielded a mean distance of 8.17 m with a confidence interval of ±1.05 m, influenced primarily by lateral sign offset and road geometry. All tagged signs positioned within the antenna’s effective coverage area were detected during dynamic tests conducted at speeds of up to 40 km/h. The results demonstrate reliable identification within a defined operational envelope, with recognition distances primarily influenced by lateral offset and road geometry, indicating that RFID-based perception is particularly suitable for controlled or semi-controlled environments.

Tokenizing Complex Passwords Using Breadth-First Search and Dictionary Matching

Salam Al-E'mari — 2026-07-14

Despite the adoption of complex password policies, users often create passwords that follow predictable patterns involving dictionary words, numbers, and symbols. Traditional tokenization techniques used for password analysis frequently overlook or misclassify symbolic and numeric elements, resulting in incomplete strength evaluations and less effective cracking strategies. This study presents a Breadth-First Search (BFS)-based tokenization framework that systematically segments passwords into meaningful components, including words from dictionaries, numeric sequences, and symbolic tokens. The BFS algorithm examines all possible substring combinations to identify the most comprehensive segmentation path. Remaining unmatched symbols and numbers are processed in a dedicated post-analysis phase to ensure complete token representation. Experiments conducted on a real-world dataset of 100,000 passwords demonstrate that the proposed approach outperforms baseline tokenizers in terms of token coverage and segmentation accuracy, while maintaining efficient processing times. The improved tokenization results contribute to a more accurate assessment of password complexity and support the development of stronger password-cracking models. These findings emphasize the importance of structure-aware parsing methods in advancing password security analysis.

Ontologies in Hearing Impairment

Ana Paula Lopes De Abreu Ferreira — 2026-07-14

Hearing impairment requires solutions that promote semantic interoperability and clinical data integration. In this context, ontologies play a fundamental role in the formal representation and standardization of biomedical knowledge, supporting data sharing across heterogeneous health information systems. This study maps, characterizes, compares, and analyzes ontologies related to hearing impairment, aiming to identify conceptual overlaps and potential points of interoperability. An exploratory-descriptive approach was adopted, following PRISMA guidelines to identify and evaluate ontologies associated with hearing health. Based on thematic relevance and conceptual scope, the Hearing Impairment Ontology (HIO) and the Systematized Nomenclature of Medicine – Clinical Terms (SNOMED CT) were selected for comparative analysis. Ontology alignment was performed using AgreementMakerLight (AML) with a similarity threshold of 0.5. The comparison identified 39 valid correspondences among 797 classes in SNOMED CT and 495 classes in HIO, indicating a low direct semantic overlap between the ontologies. Correspondences were concentrated in established clinical concepts related to hearing loss, while differences were observed in genetic, phenotypic, and therapeutic domains. The results suggest complementarity between the ontologies and reinforce the importance of semantic alignment strategies to support interoperability among clinical and biomedical data in hearing healthcare and computational audiology.

Satellite imagery super-resolution using GANs and aerial images

Magda Alexandra Trujillo-Jiménez — 2026-07-14

Satellite imagery often suffers from limited spatial resolution and, in many cases, high acquisition costs. These factors restrict their use in applications such as urban monitoring, land management, and wildlife studies. This work proposes an AI-based super-resolution approach that leverages highresolution aerial imagery to train a Generative Adversarial Network. Specifically, the ESRGAN (Enhanced Super-Resolution Generative Adversarial Network) architecture is adapted and trained using aerial orthophotos, enabling the transfer of learned spatial representations to low-resolution satellite images. The trained model is evaluated on satellite image patches at ×2 and ×4 super-resolution scales. Performance is assessed using structural, perceptual, and chromatic metrics, including SSIMY, MS-SSIM, LPIPS and CIEDE2000. The results show clear improvements, with increased sharpness, enhanced edge definition, and consistent reconstruction of urban structures and terrain features. From a quantitative perspective, the ×2 scale achieves the best overall metric values, while the ×4 scale maintains stable and meaningful performance despite the higher reconstruction difficulty. These findings demonstrate the feasibility of transferring super-resolution capabilities from aerial images to satellite imagery, even in the presence of spectral and geometric differences between acquisition domains. Overall, this study provides a solid foundation for the development of low-cost, AI-driven satellite image super-resolution models and outlines future research directions focused on dataset expansion, domain adaptation strategies, and sensor-specific architectural improvements.

Denoising of EEG Signals in Brain–Computer Interfaces Using Extended Kalman-Filtered Recurrent High-Order Neural Networks

Enrique López-Estrada — 2026-07-14

This paper presents a Recurrent High-Order Neural Network (RHONN) for EEG denoising in brain–computer interface (BCI) applications. Trained online using the Extended Kalman Filter (EKF), the proposed approach effectively suppresses EOG, EMG, and ECG artifacts under non-stationary conditions while preserving EEG temporal structure. Experimental results show that RHONN achieves performance comparable to or better than conventional filters and deep learning models, including MLPs, RNNs, GANs, and autoencoders. A key advantage of the RHONN--EKF framework is its very low computational cost. By modeling each EEG channel with a single high-order neuron, the method reduces computational load by more than 90% compared to state-of-the-art models, making it suitable for real-time, resource-constrained BCI systems and consistent with Green AI principles.

Porting a Virtual Reality Kayak Application Across Two Head-Mounted Displays: Presence, Usability, and Attitudes Toward Tourism

Jorge González-Ortega — 2026-07-14

Tourism is an important economic activity for Chile. The use of Virtual Reality (VR) technologies have a positive impact on the reception of messages towards points of interest in tourism, influencing peoples' attitudes towards tourism and intention to visit. Building on previous research on tourism, virtual reality, and attitudes, we present a porting of a VR activity of Kayaking on a lake in Chile, to two VR devices (HTC Vive and Meta Quest 3) using the Unity Editor. Differences in perceived presence and usability of the developed software are analyzed, along with attitudes towards kayaking and tourism. Results indicate that there are no significant differences in presence, usability or attitudes between participants that used the HTC Vive or the Meta Quest 3 device. Meanwhile, the scores of presence, usability, and attitudes toward kayaking and tourism are statistically higher than the theoretical midpoint of the scales, suggesting that the virtual reality experience is associated with positive evaluations of the simulated activity. The results are analyzed and discussed in relation to the characteristics of both VR devices, as well as the methodological utility of the results.

Comparison of Tree-Based Machine Learning Models for Classification of Tuberculosis Outcomes in Brazil

Heloísa de Almeida Pereira — 2026-07-14

Tuberculosis remains a significant public health concern, recognized as a reemerging disease strongly associated with socioeconomic conditions. According to the World Health Organization, tuberculosis continues to be the leading cause of death from a single infectious agent worldwide in 2025. This study evaluates Random Forest, XGBoost, CatBoost, and LightGBM for classifying four treatment outcomes (cure, abandonment, death from tuberculosis, and death from other causes) using 53,656 epidemiological records from Minas Gerais, Brazil (2010–2024). A preprocessing pipeline was designed to handle heterogeneous and high-cardinality variables. Models were assessed with and without SMOTE balancing under 5-fold stratified cross-validation, using accuracy, weighted F1-score, and per-class recall as evaluation metrics. Without balancing, all models achieved approximately 0.75 accuracy but failed to detect minority outcomes. Experiments across multiple random seeds, feature subset analysis, and temporal validation were conducted to assess model robustness. With SMOTE, CatBoost achieved the best overall performance, with the highest cross-validation F1-score (0.7071 ± 0.0034) and improved recall for abandonment and death outcomes. Results indicate that the combination of
clinical and socioeconomic features is essential for predictive performance, and that data quality and class imbalance are the main obstacles to reliable minority-class detection in tuberculosis surveillance.

Hybrid IRS–NOMA Framework for Turbulence Resilient Underwater Visible Light Communication

Mahesh Miriyala — 2026-07-14

Underwater Visible Light Communication (UVLC) offers high data rates, low latency, and strong physical-layer security, making it as an alternative to acoustic and RF systems for real-time underwater applications. However, its performance is severely degraded by absorption, scattering, and turbulence-induced fading. To enhance link reliability, recent advances in Non-Orthogonal Multiple Access (NOMA) and Intelligent Reflecting Surfaces (IRS) are being integrated into UVLC systems. NOMA enables simultaneous multiuser access through power-domain multiplexing, improving spectral efficiency and connectivity, while IRS enhances signal strength and coverage by intelligently reflecting and directing optical beams. Motivated by these benefits, we propose a hybrid IRS–NOMA framework for UVLC to improve link reliability and mitigate turbulence effects. A composite channel model is formulated by combining direct and IRS-reflected paths under absorption, scattering, and turbulence modeled using the Exponential Generalized Gamma (EGG) distribution with pointing errors. Closed-form expressions for the outage probability and average bit error rate (ABER) are derived in terms of the Meijer-$G$ function. System performance is evaluated under various configurations, and Monte Carlo simulations validate the analytical results, demonstrating excellent match and highlighting the effectiveness of the proposed IRS–NOMA-assisted UVLC design.

A State-Space Time Series Modeling of Long-Term River Flow and Precipitation Trends across Chile (1910–2015)

Francisco Eduardo Novoa-Muñoz — 2026-07-14

This study provides a comprehensive century-scale assessment of river discharge and precipitation trends across Chile (18◦S–56◦S). We reconstruct regional monthly series from 604 flow and 831 precipitation stations using a statespace model (Durbin & Koopman, 2001) that separates level, trend and seasonal components and allows imputation via the Kalman filter/smoother. Trend detection combines parametric linear regression with non-parametric Mann–Kendall and Theil–Sen estimators. Cross-correlation analysis and decadal Hovmöller diagrams are used to characterize rainfall–discharge coupling and the spatiotemporal propagation of hydrological signals. Results show marked drying trends in northern and central Chile and wetter signals in the far south, with an approximate 1,200 km southward displacement of discharge isolines since 1950. The paper documents the statistical framework in full mathematical detail to support reproducibility.

Table of Contents September 2026

Daniel Ulises Campos Delgado — 2026-07-14

Low-Complexity Single-Layer Dual-Band Microstrip Patch Antenna for WiFi 6/6E Communication Systems

Tu Le-Tuan — 2026-07-14

This paper presents a dual-band patch antenna operating at 5.2 and 6.6 GHz for WiFi 6/6E communication systems. Unlike conventional approaches based on higher-order modes or stacked structures, which often lead to larger size and complicated structure, the proposed design employs two closely spaced patches of different sizes printed on the same layer to realize dual-band operation. These patches are capacitively excited through a centrally located probe-fed microstrip line, enabling a single-layer, low-profile, and low-complexity configuration. The optimized antenna occupies an overall size of 0.52 λ × 0.52 λ × 0.03 λ at the lowest operating frequency of 5.2 GHz. The measured impedance bandwidths are 6.1% (5.12–5.44 GHz) and 4.9% (6.40–6.72 GHz) for the lower and upper bands, respectively. Within these bands, the antenna achieves maximum broadside gains of 7.0 and 6.7 dBi, while maintaining stable and desirable radiation characteristics across both operating frequencies.

Automated Design and Miniaturization of RF Resonators Based on the Cesàro Fractal for Chipless RFID Tags

Gabriel São Martinho da Silva — 2026-07-14

This paper presents an automated methodology for the design and miniaturization of novel resonators based on the modified Cesàro fractal. The resonators are generated using Python scripts integrated with the HFSS electromagnetic simulator and offer a compact alternative to conventional square-loop resonators for chipless RFID tags. A key advantage of the proposed approach is the ability to iteratively miniaturize the structures, resulting in a significantly reduced tag footprint. The Cesàro fractal, an adaptable derivative of the Von Koch curve, is explored here by treating the traditionally fixed 60° deformation angle as an adjustable parameter. This flexibility enables precise geometric optimization of the resonator for operation at target frequencies. The methodology encompasses theoretical analysis of fractal parameters such as the angle and fractal order. The script-based geometry generation, full-wave electromagnetic simulation, device fabrication, and experimental validation through comparison of simulated and measured results. The use of the Cesàro fractal is justified not only by its inherent miniaturization capability but also by the finer control it provides over the resonant frequency. We demonstrate that varying the structure’s perimeter allows for precise tuning of the resonant frequency without incurring an undesirable increase in resonator area. When a resonator is detuned, no energy coupling occurs, resulting in a high signal level at the output port, corresponding to a logical “1” state, which is essential for chipless RFID encoding. Experimental results confirm the efficiency of the Cesàro fractal as a viable alternative to conventional square-loop resonators, highlighting its potential for compact, high-performance RFID systems.

Electric Field Redistribution Driven SEGR Mitigation in a High-k Trench VDMOS with Floating Islands

Sanjeev Manoj Ranjan — 2026-07-14

Single-event gate rupture (SEGR) in trench-based power MOSFETs originates from transient oxide-normal electric fields generated during heavy-ion–induced charge transport, rather than from static breakdown limitations alone. Conventional mitigation approaches, such as stacked high-k gate dielectrics or drift-region field modulation using floating islands, address different aspects of the electric-field distribution but are typically examined independently. In this work, a TCAD-based study is presented to investigate how the co-integration of a high-k/SiO₂ stacked trench dielectric and P-type floating islands in the drift region modifies electric-field partitioning under both steady-state blocking and transient heavy-ion irradiation. The combined effect of dielectric field redistribution and drift-region charge modulation reduces the peak oxide-normal electric field and delays the onset of gate rupture. Calibrated simulations indicate an increase in breakdown voltage relative to a conventional trench Vertical double-diffused MOSFETs (VDMOS) while maintaining comparable specific on-resistance. Under heavy-ion irradiation, the reduced oxide field enables sustained operation up to a linear energy transfer of 40 MeV · cm²/mg at 35% of the rated breakdown voltage. These results clarify the role of coupled dielectric and drift-region field engineering in SEGR mitigation and provide physics-based guidance for radiation-tolerant silicon power MOSFET design.

A Quasi-LPV Dynamic Output Feedback Stabilizer for Nonlinear Descriptor Systems via Convex Optimization Techniques

Arturo Alvarado — 2026-07-14

This work proposes stabilizing descriptor systems via an output feedback controller belonging to the dynamic category, that is, observer-based controllers. The proposed approach allows for handling nonlinear descriptor systems whose non-constant terms might depend on unavailable signals. The designing conditions are linear matrix inequalities obtained from applying convex representations in combination with the Lyapunov method. Numerical examples and real-time experiments in the well-known rotatory inverted pendulum illustrate the effectiveness of the methodology.

Cascade Pixel Transformer with Distance-Driven Spatial Fusion for Hyperspectral Image Classification Using Limited Training Samples

Biri Chanakya Reddy — 2026-07-14

Classification of hyperspectral images is a crucial component of remote sensing, as it facilitates characterization of land surfaces. The limited availability of labeled data necessitates the design of architectures that can train with a minimal number of samples. Existing pixel- and patch-based methods advanced the field; however, they struggle to provide optimal results with limited training samples due to architectural limitations in learning complex spectral features and spatial contextual information from a small number of samples. To overcome these limitations, the Cascade Pixel Transformer Network (CPTNet) is proposed in this paper. CPTNet introduces a novel pixel-level transformer architecture that comprises transformer units in a cascade configuration to capture long-range spectral features. This work proposes an innovative input feature embedding strategy that combines PCA with a fully connected layer with pruned weights. CPTNet includes a distance-driven spatial fusion block that effectively aggregates the spectral features of all neighboring pixels in the input patch, which results in efficient spatial contextual information usage without relying on overfitting spatial patterns. Extensive experiments involving three benchmark datasets demonstrated the superiority of CPTNet as compared to state-of-the-art methods. The proposed CPTNet improves overall accuracies for three benchmark datasets by 3.31%, 1.71%, and 0.84% compared to the respective second-best approaches. The work proposed in the paper advances the field of hyperspectral classification by robust integration of PCA, weight pruning, pixel-level transformer units, and a distance-driven spatial fusion strategy. The source code repository for CPTNet can be found in https://github.com/profCRB/CPTNet.

High Performance 84-Pulse Driver for Speed Control of Permanent Magnet Synchronous Motors

Antonio Valderrabano-Gonzalez — 2026-07-14

Permanent Magnet Synchronous Motors (PMSMs) are widely employed in high-performance industrial On this paper we mention that Permanent Magnet Synchronous Motors (PMSMs) are widely employed in high-performance industrial applications due to their high efficiency, high power density, and excellent controllability. The overall performance of PMSM drive systems is strongly influenced by the interaction between the control strategy and the power electronic converter. This paper presents a high-performance 84-pulse voltage source converter (VSC) integrated with an advanced speed control strategy for precise PMSM regulation. The proposed approach leverages the increased voltage resolution of the 84-pulse topology to improve dynamic response, reduce torque and voltage ripple, and enhance speed-tracking accuracy under variable operating conditions. A simulation-based design of experiments is conducted to evaluate the combined behavior of the controller and converter across a wide range of speed and load torque profiles. The results demonstrate that, compared with conventional lower-pulse converter topologies, the proposed configuration achieves smoother electromechanical behavior, reduced harmonic distortion, and improved tracking performance, making it well suited for high-precision PMSM drive applications.

Enhancing the Voltage Profile and Net Profit of Reconfigured Distribution Networks with the Optimal Integration of EVCS and Capacitor Banks with Chaotic HBA

Madhubabu Thiruveedula — 2026-07-14

Nowadays, plug-in electric vehicles (PEVs) are increasing very fast because people are worried about pollution caused by petrol and diesel vehicles. As the number of electric vehicles (EVs) is growing quickly, there is a need to provide proper and efficient charging stations. But installing many charging stations creates some problems in the distribution system, such as voltage becoming unstable and increase in power losses. In this paper, a complete optimization method is proposed to improve the voltage level and increase overall profit in radial distribution networks (RDNs). This is done by placing electric vehicle charging stations (EVCS), capacitor banks, and by changing the network connections (network reconfiguration) together at the same time. In this work, a Chaotic Honey Badger Algorithm (CHBA) is used to solve the complex optimization problem which has nonlinear and mixed-integer variables. By adding chaotic mapping, the search ability of the algorithm improves and it helps to avoid getting stuck at a wrong early solution. The objective function considers both technical and economic factors. It aims to reduce power loss, improve voltage condition, decrease installation and operating cost, and increase the benefit by reducing energy losses. The proposed method is tested on the IEEE 33-bus test system under different conditions. The results show good improvement in voltage level, reduction in power losses, and increase in net profit when compared with other existing methods like PSO, GWO, and AGWOPSO. The CHBA method behaves steadily and yields consistent results under various test situations, according to the statistical analysis. The results obtained indicate that the recommended strategy is appropriate for practical implementation in present-day power distribution systems, especially in networks where electric vehicles (EVs) are highly prevalent.

Effect on the driving range of an electric vehicle when photovoltaic systems are integrated into its bodywork

Wilson Silva — 2026-07-14

Maintaining a sufficient driving range in electric vehicles (EVs) typically requires large battery capacities, which increases weight, cost, and reduces efficiency. This study evaluates the impact of integrating photovoltaic (PV) systems into the vehicle body through a comprehensive approach that jointly considers energy generation, thermal effects due to forced convection, aerodynamic losses, and annual energy performance. A mixed quasi-experimental methodology was implemented, combining experimental I–V characterization under real operating conditions with simulations based on a 3D vehicle model. Results show that the PV system can generate between 2.52 and 4.43 kWh/day, leading to a driving range increase between 3.61% and 7.81% (up to 23.42 km/day). Additionally, annual energy generation ranges from 1021 to 1616 kWh, with potential savings of up to 2044 USD/year. These results demonstrate that PV integration can enhance energy efficiency and reduce reliance on grid charging under real operating conditions.