Authors - Md. Nadimul Islam, Sajid-Ul Islam, Tahsina Islam Afra, Mohammad Shidujaman Abstract - Foliar diseases impact negatively on the health and productivity of mango trees, hence it is essential to manage them effectively. The proposed research is an automated approach to diagnosing popular in common mango leaf diseases, such as Anthracnose, Bacterial Canker, and Powdery Mildew, utilizing high-throughput imagery. The suggested methodology deploys a Transfer Learning model which employs MobileNetV2 framework which is already trained using ImageNet to guarantee successful and precise classification on battery limited devices such as Raspberry Pi. With the combination of target feature detection and a specialized classification head, the system offers real-time detection that can be used in spraying mechanisms using the IoT. Through experimental analysis, it is shown that the proposed CNN-based framework is highly accurate in terms of classification when the experiment is conducted under controlled conditions and as such, the framework has potential to be used in automated mango leaf disease detection.
Authors - Kazi Saiful Islam, Sadman Kabir, Abir Sen Gupta, Sayra Islam Saki, Md. Tafshir Jaman Takib, S.M. Sayem Abstract - This paper explores the critical role of Green Innovation and Green Finance Index in influencing Sustainable Business Performance with a specific focus on Green IT Capital as mediator. For primary data collection, questionnaire was distributed among Bangladeshi employees appointed in several industries and 407 responses were obtained. The Partial Least Square Structural Equation Modelling (PLS-SEM) approach was used for the data analysis. The findings demonstrate that Green Innovation (consisted of Green Product Innovation, Green Process Innovation and Green Technology Innovation) as well as Green Finance Index (consisted of Green Bond and Green Investment) positively influence Sustainable Business Performance. Moreover, Green IT Capital directly impacts Sustainable Business Performance. Additionally, Green IT Capital significantly mediates the relationship of Green Finance Index and Sustainable Business Performance, however, significant mediation between the relationship of Green Innovation and Sustainable Business Performance was not found, which is a central finding of this study. The results infer several insights for firms to utilize the funds to integrate Green IT Capital in their core activities to attain sustainable outcomes. The findings clarify the need to arrange policies to incentivize Green IT Capital adoption across industries. These factors may enhance Green Communication strategies and accelerate the nation to attain SDG 9 and SDG 12.
Authors - Thinh Truong, Chau Vo, Anh Duong Abstract - The early diagnosis of patient conditions at the hospital admission stage is crucial for optimizing medical resource allocation, reducing overcrowding, and improving patient outcomes. Traditional diagnostic approaches at admission rely on limited initial information and expert assessment, which can lead to misclassification and delayed treatment. This paper proposes a multimodal data-driven approach that integrates Large Language Model (LLM) to predict patient conditions using structured and unstructured medical data. In particular, we propose a classification model that leverages LLM for multimodal data processing and generates feature representation based on demographics, biometrics, vital signs, lab values and electrocardiogram (ECG) data for 78-disease diagnoses. Compared to the existing models, our model decides a better data fusion with semantics-preserving. Indeed, evaluated through experiments on the constructed dataset from MIMIC-IV using standard metrics such as Area Under the Receiver Operating Characteristic (AUROC), Precision, Recall, and F1-score, the proposed model outperforms traditional ones. Experimental results also highlight the potential of integrating multiple data sources for automated patient triage at the admission stage.
Authors - Kiwa Matsui, Teruki Toya, Kenji Ozawa Abstract - Medical percussion estimates internal body conditions from acoustic responses generated by tapping the body. To enable portable and comfortable health monitoring, this study proposes a music-based electronic percussion system using ordinary musical signals as test sig nals. The system improves portability by introducing a compact piezo speaker exciter and a vibration pickup fixed to an abdominal band. In ad dition, signal safeguarding is applied so that musical signals can be used for impulse-response measurements with sufficient spectral power. Exper iments measuring stomach responses before and after meals showed that the safeguarded musical signals produced results comparable to sweep signals and enabled detection of state changes. These results demon strate the feasibility of portable, noninvasive health monitoring using music-based electronic percussion. Furthermore, arbitrary music signals can be converted into reliable excitation signals through signal safeguard ing while preserving perceptual musical quality.
Authors - Ajinkya Chavan Abstract - The pursuit of intelligent systems capable of parsing human intent and navigating complex information landscapes has evolved from rigid, rule-based architectures to sophisticated, agentic frameworks. Early prototypes, such as the "Artificially Talented Architecture" (ATA), demonstrated the foundational utility of theme detection coupled with rudimentary holographic interfaces; however, these systems were constrained by the independence assumptions of Vector Space Models (VSM), limited context windows, and a lack of semantic relationship modeling. In the current era of Generative AI, while Large Language Models (LLMs) have solved fluency, they continue to struggle with "Global Sensemaking"—the ability to synthesize highlevel themes across vast corpora without succumbing to hallucination or context fragmentation. This paper introduces Holo-Agentic GraphRAG, a novel architecture that integrates Agentic Retrieval-Augmented Generation (Agentic RAG) with spatial computing to redefine state-ofthe- art theme detection. Unlike traditional methods relying on flat retrieval, the proposed approach employs a hierarchical knowledge graph constructed via LLM extraction and refined through the Leiden community detection algorithm. This structure allows for dynamic graph traversal and multi-level summarization. Furthermore, user interaction is formalized as a Partially Observable Markov Decision Process (POMDP) within a mixed-reality environment, fusing gaze tracking and voice prosody to resolve communicative ambiguity. Experimental results on the GraphRAG-Bench and a proprietary spatial interaction dataset demonstrate that Holo- Agentic GraphRAG outperforms standard RAG and static GraphRAG baselines by 18.4% in multi-hop reasoning accuracy and 22% in theme detection coherence, while significantly reducing token overhead.
Authors - Saki Matsudo, Yurika Obata, Koichiro Kido, Kenji Ozawa Abstract - We present First-B/M, a smartphone-based system that enables pregnant women to measure fetal heart rate (FHR) at home, analogous to auscultation. The system integrates two external microphones embedded in short stethoscope tubes, connected to an iOS device. The smartphone performs low-pass ltering (250 Hz), harmonic/percussive sound separation (HPSS) to extract fetal heart sounds, and FHR es- timation based on frame-wise amplitude increases. To improve robust- ness in non-clinical environments, a median-based temporal aggregation method is applied. A user-centered application supports recording, FHR visualization, and data sharing with clinicians. Usability was assessed with 10 participants through task completion and ve-point rating evaluations. System performance was evaluated using 20 recordings from pregnant women, in which fetal heart sounds were identiable in 12 cases. Supplementary pseudo-fetal data, generated by time-scaling adult heart sounds, were used to examine algorithm behavior under ideal con- ditions. When fetal heart sounds were captured, estimated FHR values agreed with human reference measurements within 3%. One outlier occurred under strong mid-recording noise, indicating the need for auto- matic re-measurement support. These results demonstrate the feasibility of smartphone-based auscultation for home FHR monitoring and provide a practical foundation for non-clinical FHR measurement systems.
Authors - Ranjan Kumar Behera, S. Dinesh Naveen Kumar Abstract - Social networking platforms such as Twitter have become inuential spaces where users routinely express opinions, emotions, and personal experiences providing valuable signals for understanding mental health conditions. This study leverages such user generated con- tent to investigate depression indicators and analyze their prevalence across countries classied as developed and developing. Unlike traditional sentiment analysis approaches, this work introduces a novel attention enhanced BiLSTM architecture combined with a hybrid ensemble framework specically tailored for depression detection in short, informal social-media text. The proposed model integrates contextual attention with bidirectional sequence learning to capture subtle linguistic cues, while the ensemble mechanism enhances robustness against noise and linguistic variability across regions. The proposed methodology involves a comprehensive preprocessing pipeline, depression-lexicon construction, machine-learning baselines, and the proposed deep model. Experimental evaluation demonstrates a signicant improvement in detection accuracy and generalization, out performing existing benchmark methods. The study also presents a unique cross-country comparative analysis of de- pression trends, o ering insights into how socio-economic environments in uence online emotional expression.
Authors - Sumet Jirattisak, Tanatorn Tanantong, Nittaya Chemkomnerd Abstract - Psoriasis is a chronic autoimmune skin disease, and accurate diagnosis remains challenging due to the shortage of dermatologists and the subjective na ture of visual assessment. To address this challenge, this study developed an au tomated classification system using three deep learning architectures, Efficient Net-B4, MobileNetV3, and Vision Transformer, within a transfer learning frame work to classify Psoriasis, Healthy Skin, and Psoriasis-like Disorder images. The models were fine-tuned and evaluated using 5-fold cross-validation on three da tasets: the Thammasat University Hospital dataset, the Kaggle dataset, and a combined dataset derived from DermNet and a previously published study in volving Indian patients. EfficientNet-B4 achieved the highest accuracy on the TUH dataset (99.68%) and the Dermnet-India dataset (94.40%), while Mo bileNetV3 performed best on the Kaggle dataset (96.88%) and required the short est training time. Overall, the results show that EfficientNet-B4 offers superior predictive performance, whereas MobileNetV3 provides a better balance be tween accuracy and computational efficiency. The findings confirm that transfer learning is a time-efficient approach for psoriasis classification, reducing training time and computational cost while maintaining acceptable performance, particu larly under limited clinical data conditions.
Authors - Sangeeta Singha, Lalhriatpuii, Banani Basu, Arnab Nandi Abstract - This research presents a microwave-based breast cancer detection framework that leverages the Specific Absorption Rate (SAR) of an Ultra-Wideband (UWB) patch antenna, operating between 3.1 and 10.6 GHz. By positioning an antenna array on opposite sides of a breast phantom and rotating it, the system records SAR distributions as 2D input images. To isolate pathological features, image segmentation is performed on these 2D data samples to distinguish between healthy, benign and malignant tissue. These processed images are then classified using a ResNeXt architecture integrated with a Spatial Attention Module (SAM) to enhance tumor detection. Experimental results demonstrate the efficacy of this attention-driven approach, as the integration of the SAM improved classification accuracy to 98.44%.
Authors - Stefan Lippitsch, Mario Hirz Abstract - Automated conductive charging of electric vehicles using robotics can increase availability and user convenience, especially in depot and fleet applica tions. At the same time, new safety-critical situations arise from close human robot-vehicle interaction, changing environmental conditions and the coupling between charging infrastructure and electric passenger cars. This paper presents a camera-based robotic system for automated conductive charging with standard ized connectors, including the overall system architecture, perception for detect ing the charging flap and standardized charging inlet, robust pose estimation and a state-based process control. The second part introduces a framework developed to perform a hazard anal ysis and risk assessment specifically tailored to automated charging processes. The approach includes a discussion of relevant (functional) safety standards from machinery and robotics domains and their applicability to automated charging, linking functional safety with general machine and collaborative robotics safety. Additionally, an evaluation method is introduced, enabling a traceable deriva tion of safety goals for this use case. Finally, a comparison is made to the auto motive equivalent functional safety standard using performance parameters. The presented methodology supports consistent risk reasoning across disci plines and provides a practical foundation for developing scalable, compliant, and risk-optimized automated charging systems.
Authors - Jay Joshi, Avneesh Jadhav, Ishita Deshpande, Ameya Dharap, D. D. Sapkal Abstract - As medical insurance adoption continues to grow and its complexities continue to increase, insured members require trustworthy and clear guidance, transparency and timely progress updates throughout the insurance lifecycle. However, users often run into fragmented information, confusion in policy selection, incomprehensible policy documents due to tremendous technical jargon and limited procedural guidance. This makes it difficult to understand coverage details and navigate claims smoothly; particularly during medical emergencies. The absence of unified communication channels frequently leaves policyholders uncertain about eligibility, documentation requirements and claim progress, leading to stress and reduced trust in insurance services. This paper proposes a user-centric, AI-enabled digital platform designed to improve transparency and communication between insured members and insurance service providers. The system focuses on simplifying policy discovery through personalized policy recommendations and interpretation through NLPbased clause summarization. These features enable users to gain a clear understanding of inclusions and exclusions, which help them to make informed decisions. Additionally, to support users during claims, the RAG-based assistance module provides step-by-step guidance on eligibility, document submission and claim procedures. By emphasizing clarity, continuous guidance and transparency, the proposed solution enhances user experience, reduces claim-related anxiety and encourages trust and adoption of digital healthcare insurance services.
Authors - U.H.S. Rashmina Amarasinghe, K.A Dilini T. Kulawansa Abstract - This literature review examines the expanding and critical role of Artificial Intelligence, including Machine Learning and Deep Learning, in countering increasingly complex cyber threats. The purpose of this review is to analyze the applications, effectiveness, challenges, and future research directions of Artificial Intelligence driven technologies in threat detection. Artificial Intelligence driven systems significantly enhance the NIST Cybersecurity Framework functions (Identify, Protect, Detect, Respond, Recover). They excel at real time anomaly detection in massive datasets, outperforming traditional signature-based methods against modern attacks like zero-day exploits and polymorphic mal-ware. Key techniques discussed include Support Vector Machines, Decision Trees, and various Neural Networks used in effective Intrusion Detection Systems and phishing classification. However, the review highlights the dual nature of Artificial Intelligence, noting the rise of Artificial Intelligence driven cyberattacks and the challenges posed by high resource demands and managing data quality. Ethical considerations, specifically concerning privacy and transparency, necessitate the development of Explainable Artificial Intelligence. Ultimately, the future relies on Hybrid Augmented Intelligence, a strong human, Artificial Intelligence collaboration to maintain effective cyber defenses.
Authors - H.M.H.H. Gunarathne, K.A. Dilini Kulawansa Abstract - Federated Learning enables the collaborative development of AI models in healthcare while preserving patient data confidentiality, offering a promising solution to privacy, regulatory, and data transfer challenges. Unlike conventional centralized learning, FL transmits only model updates, including gradients or aggregated parameters, rather than raw data, thereby enabling multiple institutions to collaboratively train models while maintaining data confidentiality. This review outlines that FL ensures model accuracy and generalizability of the model in privacy-aware healthcare applications. It also discusses more privacy preservation methods that are implemented in combination with Federated Learning, including Differential Privacy, Homomorphic Encryption, Secure Multi-Party Computation, and blockchain-based systems, which help to increase security, trust, and transparency. The paper has also reviewed the existing studies in the key areas of healthcare such as disease diagnosis, medical im-aging, remote patient monitoring, predictive analytics and Electronic Health Record management. By demonstrating the potential of FL to enable scalable, secure, and privacy-preserving AI systems, this review provides insights into its transformative role in advancing intelligent, patient-centered healthcare solutions.
Authors - D.M. Jarathne, K. A. Dilini T. Kulawansa Abstract - The cryptographic systems underlying the digital infrastructure of the world present an existential risk to quantum computing. With wide deployment of cryptographically relevant quantum computers, many commonly deployed asymmetric encryption algorithms including RSA and elliptic curve cryptography will be subject to attack through quantum algorithms such as the Shor algorithm. The present systematic literature review examines the feasible and scalable migration plans to deploy enterprise systems and critical infrastructure to post-quantum cryptography. The review explores migration frameworks, implementation issues, practical implementation, and organization strategic recommendations based on the analysis of fifteen selected sources, including research articles and technical standards. The review notes that there are four basic stages of migration which include diagnosis, planning, execution, and maintenance. No-table obstacles are organizational issues, technological constraints, system over-load, and industry-specific demands. Practical examples of successful migrations between web servers, databases, blockchain architectures, and messaging systems have been reported, and hybrid cryptographic solutions have become the most common transitional practice.
Authors - Nailfaaz, Wahyono Abstract - Accurate segmentation of anatomical structures in chest radiography (CXR) is critical for automated diagnosis. While CNNs achieve high regional overlap, they struggle with precise organ boundaries due to X-ray projection artifacts. This study systematically evaluates 32 encoder–decoder configurations combining U-Net and DeepLabV3+ with ResNet, MobileNet, and EfficientNet families to isolate Conditional Random Field-as-RNN (CRF-as-RNN) refinement impact on boundary quality. Results show U-Net outperforms DeepLabV3+ in preserving anatomical details. Crucially, a ”capacity threshold” is identified: CRF integration significantly reduces Hausdorff distances for lightweight models but yields diminishing returns for high-capacity backbones where baseline topology is already optimal.
