Abstract

Tuberculosis (TB) remains a leading infectious cause of death worldwide, and timely, accurate clinical decision-making is essential for diagnosis, treatment selection, and management of drug-resistant strains. Conventional training methods often leave healthcare providers underprepared for the nuanced, high-stakes decisions TB care demands. Mobile application-based case simulation platforms address this gap by delivering interactive, branching clinical scenarios directly to smartphones, allowing learners to practice diagnostic reasoning, order and interpret investigations, and select treatment regimens with real-time, algorithm-driven feedback. Such platforms support self-paced, repeatable learning and are particularly valuable in resource-limited, high-burden settings where access to expert mentorship and structured training is scarce. Preliminary evidence indicates that mobile case-based learning improves diagnostic accuracy, adherence to treatment guidelines, and learner confidence among medical students, nurses, and community health workers. Their offline functionality and low bandwidth requirements further enhance scalability across varied geographic contexts. Embedding mobile-based digital case simulations into TB training curricula offers a practical, cost-effective strategy to strengthen clinical competencies and advance global efforts toward TB elimination.

Biography

Dr. Arvind M. Jagtap is an accomplished academician, researcher, and mentor with over 18+ years of experience in teaching, research, and institutional development. Currently serving as a Head & Associate Professor at Computer Engineering Department, Vidya Pratishtha’s Kamalnayan Bajaj Institute of Engineering and Technology (VPKBIET), Baramati, he plays a pivotal role in nurturing a research-driven academic culture and building strong academia-industry linkages. His areas of expertise include Artificial Intelligence (AI), Machine Learning (ML), data analytics, and curriculum innovation. Dr. Jagtap is passionate about applying AI/ML for solving societal problems aligned with the Sustainable Development Goals (SDGs) and has actively led projects, workshops, and student innovation programs in this space. He has 26+ research publications, 11 patents, and 32 copyrights to his credit.

Abstract
In the preliminary design phase of medium-altitude long-endurance (MALE) unmanned aerial vehicles (UAVs), achieving reliable yet computationally inexpensive aerodynamic prediction remains a critical bottleneck, primarily due to the severe scarcity of available experimental wind-tunnel data. To address this challenge, the present paper systematically benchmarks three widely accessible numerical methods characterized by different levels of fidelity, applying them to a standard Heron-type UAV configuration. The evaluation encompasses a high-fidelity Reynolds-averaged computational fluid dynamics (CFD) solver implemented via ANSYS Fluent, a medium-fidelity semi-empirical component build-up code utilizing Missile DATCOM, and a low-fidelity vortex-lattice tool leveraging OpenVSP. These three distinct methodologies are applied to identical vehicle geometry across a comprehensive angle-of-attack range spanning from 0 to 22 degrees, evaluated at two distinct free-stream velocities of 20 m/s and 60 m/s. Their respective predictions for lift, drag, and overall aerodynamic efficiency are cross-verified through rigorous quantitative analysis. To ensure the reliability of the reference data, a comprehensive mesh-independence study was conducted, establishing that a computational grid comprising approximately 3.3 million cells offers an optimal compromise between numerical accuracy and computational cost for the high-fidelity CFD model. Taking the verified CFD results as the baseline reference, the low-order simulation tools show acceptable agreement within a 10 per cent margin over the mid-range of incidence angles, while consistently predicting the critical aerodynamic stall angle at approximately 12 to 13 degrees. However, significant discrepancies are observed at the extremes; the low-order tools diverge by up to 47 per cent in lift prediction at zero incidence and up to 39 per cent in drag prediction near the stall regime. This deviation is chiefly attributed to the inherent two-dimensional geometric simplifications adopted in the semi-empirical model. Furthermore, the investigation reveals that the free-stream speed exerts a negligible influence on the aerodynamic coefficients over the tested subsonic range. Ultimately, practical guidance is formulated to assist engineers in selecting the most appropriate tool depending on the specific phase of aircraft design.

Biography
Nguyen Thi Dieu Linh, Ph.D. is a Deputy Head of Science and Technology Department at Hanoi University of Industry, Hanoi, Vietnam. With expertise in Electronics, IoT, Smart Gardens, and Telecommunications, they have contributed significantly to academic research and global scientific exchange, possessing over 20 years of experience. She has authored numerous research articles and served as an Editor for several prominent books published by CRC Press and Springer Singapore, including the “Intelligent System and Networks” series indexed by Scopus. Additionally, she serves as an editorial board member for various reputed international journals, such as the Mobile Networks and Application Journal. As a prominent academic leader, she has been invited to numerous international conferences as a Keynote Speaker, Session Chair, and Guest of Honor. Their research interests include Artificial Intelligence, Machine Learning, Cyber Security, Data Science, Medical Informatics, and Software Reliability. Furthermore, she has demonstrated exceptional leadership as the Organizing Chair for major technical events, notably the International Conference on Intelligent Systems & Networks (ICISN 2021-2026) and the International Conference on Research in Management and Technovation (ICRMAT 2021-2026).

Abstract

Computer Vision has revolutionized the way machines can perceive, analyze and interpret the visual data, and biometric technologies have become basic for safe and intelligent human identification. This invited speech will discuss the importance of biometrics in the Computer Vision field and how the development of image processing, pattern recognition, deep learning, and artificial intelligence has contributed to the improvement of biometric systems not only in terms of accuracy but also of robustness and scalability. The presentation introduces the main modalities of biometrics such as face, iris, fingerprint, palmprint, gait and multimodal biometrics as well as their underlying computer vision techniques in feature extraction, representation learning, and recognition. It also talks about the recent advancements that are powered by Convolutional neural networks, Vision transformers, generative AI, explainable AI, and multi-agent intelligence, which are making adaptive, trustworthy, and privacy-preserving biometric solutions possible. Other challenges like illumination variations, pose changes, occlusion, aging, presentation attacks, dataset bias, fairness and privacy preservation are also explored, and solutions recently proposed to address these challenges are presented. Lastly, the talk discusses the recent advances in research in the areas such as smart surveillance, healthcare, digital identity, border security, financial services, and intelligent man-machine interaction, and looks forward to future trends for secure, explainable and ethical biometric systems. The aim of the session is to give the researchers, academicians and industry professionals a detailed knowledge about the ongoing progress of Biometrics in the field of current Computer Vision and intelligent decision making systems.

Biography

Dr. Juhi Chaudhary is an Assistant Professor in the Department of Computer Science and Engineering at Shri Ram Murti Smarak College of Engineering & Technology (SRMS CET), Bareilly, India. She earned her Ph.D. in Computer Science and Engineering from Guru Gobind Singh Indraprastha University (Delhi State University), New Delhi, India in 2025, where she pursued her doctoral research as a UGC Junior Research Fellow (JRF). She also qualified both the UGC-NET Junior Research Fellowship (JRF) and UGC-NET for Assistant Professor in Computer Science & Applications in 2019. With over 11 years of teaching and research experience, Dr. Chaudhary’s research focuses on Computer Vision, Digital Image Processing, Pattern Recognition, Biometrics, Face Recognition, and Intelligent Vision Systems. Her research has made significant contributions to developing robust face recognition techniques under challenging real-world conditions, particularly addressing variations in illumination and feature representation. Dr. Chaudhary has authored and co-authored more than 15 research publications, including papers in SCIE, Scopus, and Web of Science indexed journals, Springer book chapters, IEEE conference proceedings, and international conferences. She is also the inventor/co-inventor of two patents, including an Australian granted patent and an Indian published patent. Her work has received recognition within the research community for its novelty and practical applications in biometric authentication and computer vision. She actively serves as a reviewer for several reputed SCI and Scopus indexed journals published by leading international publishers, including Springer, Elsevier, Wiley etc. She is currently actively involved in academic coordination, student mentoring, and technical club activities, contributing to holistic student development. She has actively participated in numerous Faculty Development Programs, workshops, and training sessions organized by premier institutions such as IITs, NITs, and national bodies like AICTE, focusing on emerging areas including Artificial Intelligence, Machine Learning, Data Science, etc.

Abstract

 Accurate ECG arrhythmia classification across unseen clinical cohorts remains a significant challenge due to domain shifts caused by differences in patient populations, acquisition devices, and recording conditions. To address this issue, we propose a novel domain-generalized framework, termed Cross-Domain Spherical Alignment and Prototype Consistency Learning (CDSA-PC), for robust ECG arrhythmia classification. The proposed method combines cross-domain feature alignment on a hyperspherical representation space with prototype consistency regularization to learn domain-invariant and transferable ECG representations. In addition, a Physiological Squeeze-and-Excitation Residual Network is employed to capture both temporal and morphological characteristics of ECG signals.

Experimental evaluations conducted under a leave-one-domain-out setting across multiple heterogeneous ECG datasets demonstrate the effectiveness of the proposed framework. The model achieved average Accuracy, Macro-F1, AUROC, and AUPRC scores of 95.3%, 92.2%, 97.1%, and 89.2%, respectively. Ablation studies further confirm the complementary contributions of the proposed alignment and prototype consistency mechanisms. The results indicate that CDSA-PC provides a robust and generalizable solution for ECG arrhythmia classification in cross-domain clinical environments and has the potential to improve the reliability of AI-assisted cardiovascular diagnosis.

Keywords:

ECG     Arrhythmia       Classification;   Domain Generalization; Cross-Domain   Learning; Representation Learning; Deep Learning; Cardiovascular AI.

Biography

Sarvin Chitsaz is a researcher affiliated with Arak University, Iran. Her research interests encompass Artificial Intelligence, Machine Learning, Deep Learning, and Computer Vision. She is particularly interested in developing intelligent data-driven systems and exploring the application of AI techniques to address real-world challenges. Her current research focuses on machine learning and deep learning methodologies for data analysis, predictive modeling, and intelligent decision support systems.

Abstract

Effective drafting of construction contract conditions is essential for minimizing disputes and ensuring efficient project delivery. While prior studies have explored methods for selecting suitable contract forms, limited attention has been given to comparing multiple forms to identify missing clauses. This research proposes an integrated NLP-based framework that supports both contract form selection and completeness assessment. Guided topic modeling is used to evaluate major contractual criteria from literature across different standard forms, user-defined priorities are incorporated through AHP and TOPSIS to determine the most suitable contract conditions form. A pretrained Legal-BERT model is applied to compare the chosen form against others, detecting missing clauses through semantic similarity analysis. Multiple similarity measures, including Cosine, Manhattan, and Jaccard, are tested, with sensitivity analysis to identify optimal thresholds. Results show that Cosine similarity achieves the highest performance, with an F1 score of 91.8%, demonstrating the framework’s effectiveness in enhancing automated contract conditions assessment.

 Biography

Mohab Nasser is an Assistant Lecturer in Construction Engineering and Management at Cairo University and a practicing contracts engineer specializing in contract analysis, cost estimation, and procurement evaluation. His research focuses on applying artificial intelligence and natural language processing to construction contract analysis and decision-making. He has publications mainly on automated contract clause classification and AI-supported selection of construction contract types.

Abstract

This work presents an alternative method for the inverse kinematics problem when serial chain industrial robots are considered. Such method is based on an iterative search algorithm that solves the inverse kinematics by allowing that only one joint to change the robot movement at a time, in a simulation stage. This transforms the n-dimensional problem in simpler one-dimensional problems, which analytical solution for each joint is presented employing the Denavit-Hartenberg convention. The considered method does not possess internal singularities. Furthermore, due to its simplicity, it is shown experimental results obtained employing a 3-Degrees-of-Freedom redundant manipulator.

Biography

Breno S. Nucci holds a degree in Industrial Mechatronics Technology from the Faculty of Technology of Mogi Mirim “Arthur de Azevedo” (2022), a Master’s degree in Electrical Engineering from the State University of Campinas (UNICAMP) (2024), and is currently pursuing a PhD in Electrical Engineering at the same institution. He is a member of the research group at the Laboratory of Modular Robotic Systems (LSMR), affiliated with the School of Electrical and Computer Engineering at UNICAMP, with experience in electrical engineering and a focus on industrial robotics, working mainly in the areas of robotics, nonlinear systems control, neural networks, mobile robotics, trajectory control, and robotic joint control.

Abstract

As systems become increasingly interconnected, traditional cybersecurity threat identification strategies often fail to defend against nefarious actors and their sophisticated attacks. Given the current internet protocols (IP) meant to protect against email infiltrations such as Domain-based Message Authentication, Reporting and Conformance (DMARC), DomainKeys Identified Mail (DKIM), and Sender Policy Framework (SPF), a more comprehensive security analysis is desired for early threat identification. These protocols are intended to address specific email-based threats, however; the threat landscape being remediated is not typically incorporated into system conceptual or logical models. This presentation demonstrates a model-based systems engineering (MBSE) approach for threat modeling that leverages the ISO 21434 – “Road vehicles – Cybersecurity Engineering” additions to the Systems Modeling Language (SysML) profile. Capabilities of the extension include the ability to represent threats, attack trees, and hazards for further analysis. These elements are linked to structural and behavioral elements of the system architecture to maintain traceability throughout the lifecycle. Integrating a cybersecurity perspective into system structural and behavioral models promises to improve security posture of the overall architecture. There is an increase in visibility of threats to IPs by integrating security threats into a system model, increasing transparency among affected team members. Benefits to the audience include an enhanced understanding of email protocols, how they are shown in a system model, current analysis techniques, and threat mitigation methods.

Biographies

Sarah Rudder is a systems engineer and ontologist for Enola Technologies and a current PhD candidate at Colorado State University specializing in model-based systems engineering tools, languages, and approaches.