Objectives: To design a robust pancreas segmentation framework for Magnetic Resonance Imaging (MRI) images that minimizes the shift between different centers while maintaining the organs boundary and topology. Method: SOT-PanNet is a three-dimensional encoder-decoder, discrete-cosine spectral feature decomposition, Wasserstein optimal-transport barycenter alignment and topology-guided signed-distance decoding scheme. 767 multi-center MRI scans from five institutions are used in separate experiments, T1 and T2. The leave-one-center-out validation is given for unseen-center assessment, while PanSegNet and CTraTL-UNet are used for comparison. Findings: SOT-PanNet obtained Dice scores of 93.05% and 93.12% for pancreas T1-weighted and T2-weighted MRI segmentation, respectively and Jaccard scores of 87.42% and 87.35%, respectively. The accuracy was 93.76% for the T1-weighted images and 93.58% for the T2-weighted images, the recall rate was 92.18% and 91.96%, respectively. SOT-PanNet achieved 0.62 and 1.25 percentage points more on Dice score and 8.49% and 9.82% less HD95 score for T1-weighted and T2-weighted scans compared to existing CTraTL-UNet work. The comparison of segmentation performance between the different centers using leave-one-center-out method showed HD95 of 0.97 mm, 1.01 mm and ASSD of 0.20 mm, 0.21 mm from five centers. The results show an increase in volumetric overlap, boundary accuracy and robustness to variations between centers. Novelty: SOT-PanNet is a novel method that simultaneously uses spectral decomposition, optimal-transport center barycenters and topology-guided boundary decoding for center-robust pancreas MRI segmentation without adversarial discrimination. Keywords: Pancreas segmentation, Magnetic resonance imaging, Optimal transport, Spectral learning, Topology guidance
Objectives: To detect suspicious fraudulent activities in Ethereum wallet transactions using a graph-based machine learning and anomaly detection techniques. Method: From the Ethereum transaction graph, graph structure signals, timing cues, and node level attributes, have been analyzed using unsupervised models like Isolation Forest, Local Outlier Factor, and One-Class SVM to flag irregular wallet patterns. Finally, based on the transaction tracing pattern, illicit transactions are figure out for forensics actions. Findings: This work successfully found some suspicious Ethereum wallet behaviors by mixing graph analytics with unsupervised anomaly detection techniques. Graph-based signals such as PageRank, clustering coefficient, transaction frequency and entropy, boosted the capture of strange transaction behaviors that point out as suspicious operation. Isolation Forest did well at isolating the most extreme outliers, while LOF and One-Class SVM highlighted wallets with odd local density, plus a clear behavioral mismatch observed. Overall, the framework worked with the Ethereum transaction dataset from Kaggle, where wallet addresses. Out of 4000 transaction history, top 20 are figure out as may be fraudulent transaction based on their behaviors. Novelty: The study integrates graph analytics, blockchain forensics, and unsupervised machine learning to detect anomalous Ethereum wallet behavior without requiring labeled cryptocurrency fraud datasets. Here, only 0.5 percent Ethereum transaction are categorized as suspicious on this study. Keywords: Blockchain Tracing, Cryptocurrency Transactions, Ethereum, Fraud Detection, Graph Neural Network (GNN)
Background: Anxiety is a natural reaction of the body toward stress. College students frequently experience anxiety, particularly first-year nursing students who are exposed to a wide range of health care subjects and clinical postings. Preventive measures can help reduce anxiety levels among nursing students and improve their exam results. Objectives: This study aimed to evaluate the effectiveness of aromatherapy in reducing clinical examination anxiety among nursing students. Method: A true experimental two-group pre-test and post-test design with an experimental and a placebo group was adopted. Ninety-four nursing students were selected based on inclusion and exclusion criteria and were randomly assigned to either group. Baseline characteristics, pre-test anxiety levels, and biophysiological measures were collected. Aromatherapy was given to the experimental group after the pre-test. A post-test with the same tools was taken 15 minutes after the intervention. The same process was followed for the placebo group; however, instead of aromatherapy with essential oil, they were provided with three drops of normal water on a non-absorbent cloth for 15 minutes. Findings: The mean anxiety scores in the pre-test phase were 47.47 ± 4.02 (experimental group) and 45.43 ± 3.66 (placebo group) (t = 2.574, p = 0.012), indicating a statistically significant baseline difference. In the post-test phase, the mean anxiety score was 42.38 ± 3.55 (experimental group) and 48.13 ± 4.76 (placebo group) (t = –6.632, p ≤ 0.001), indicating a significant difference after the intervention and revealing that aromatherapy had a calming effect before the examination. Novelty: The article presents a placebo-controlled research evidence to support the use of lemon essential oil aromatherapy as an intervention because of its low costs and effectiveness in decreasing the number of psychological and physiological signs of feelings of anxiety during clinical examination in nursing students. Keywords: Clinical examination anxiety, Nursing Students, Aromatherapy, Lemon oil, Objective structured clinical examination
Background: Machine learning is increasingly applied to drought monitoring and forecasting, but studies differ widely in how drought is defined, what data are used, and how models are validated, which makes reported performance difficult to compare directly. Objectives: To explain why reported machine-learning drought-model performance varies across definitions, data sources and validation schemes, and to assess whether recent studies adequately address rare severe events and cross-domain transferability. Method: Journal articles published January 2018 to August 2026 were reviewed, starting from a 70-record high-relevance subset of a 359-record discovery export and extended through citation chaining. Publisher and DOI verification produced a final evidence base of 55 articles: 48 primary studies and 7 reviews. Target definition, input data, model family, validation design, metrics, class-imbalance treatment, explainability and external testing were compared across studies. Findings: Reported performance varied with the modelling task rather than with any single algorithm: Random Forest reached 97.7% AUC-ROC for agricultural-drought mapping in Queensland, SVM reached average R²=0.90 for hydrological drought in Algeria, and a CatBoost stacking model reached R²=0.9065 for remote-sensing-based drought monitoring. These scores are not directly comparable because targets, data and validation protocols differ. Recent studies increasingly test cross-product and cross-region transfer, use SHAP/LIME explainability, and adopt extreme-aware losses, yet class-wise reporting of severe and extreme drought, ordinal-aware evaluation, and leakage-resistant temporal validation remain inconsistently applied across the 48 primary studies. Novelty: The review jointly evaluates predictive accuracy alongside rare-event sensitivity, ordinal drought severity, leakage-resistant validation, explainability, uncertainty and cross-domain generalization, criteria largely absent from earlier machine-learning drought reviews. Keywords: Drought Forecasting, Machine Learning, Remote Sensing, Explainable AI, Extreme Drought
Objectives: Integrating Internet of Things (IoT) technologies in healthcare creates opportunities to develop intelligent, non-invasive, and user-friendly systems for continuous health monitoring. These advancements can simplify healthcare-related tasks, improve patient monitoring, and enable more personalized healthcare services. This study presents IntelliMirror, an IoT-based smart mirror for mood and health monitoring, which provides an interactive platform for accessing health-related information and mood-related conditions. Method: The experiment is conducted in three phases, comprising a baseline pre-survey, active IntelliMirror design and development, and interaction. Finally, the post -implementation evaluation. The experiment aims to incorporate user’s needs into the design to increase the system's trust and usability. Fifty participants completed the survey-based evaluation. The baseline survey identified user’s perceptions of the system before development, focusing on perceived health benefits, interest in mood monitoring, technological comfort, trust in data handling, and willingness to use the technology. Findings: The evaluation results show that 60% of participants perceived potential health benefits, 70% expressed interest in mood monitoring, and 62% were willing to use an intelligent mirror in a healthcare environment. The evaluation demonstrated generally favourable user perceptions, with positive responses exceeding 80% across the evaluated system features. The results indicate that IntelliMirror has promising potential as a user-centered IoT platform for integrating health and mood monitoring into an interactive smart-mirror environment. Novelty: This study explores user’s perception of an Intelligent Mirror before and after development. Based on pre-survey findings IntelliMirror was designed as a user centered IoT platform integrating healthcare and mood monitoring into a smart mirror environment. Post-development result indicates potential for smart technology adaptation in personalized healthcare. However, the research needs to include technical validation, clinical comparison, and long-term evaluation to ensure reliability and clinical applicability. Keywords: IntelliMirror, Smart Mirror, Internet of Things, IoT Healthcare, Mood Monitoring, Health Monitoring, User Acceptance, Smart Healthcare, Human–Computer Interaction
