Abstract: This scoping review of original literature published before 1 March 2025 examined the de-mographic, clinical and simple laboratory findings associated with the development of severe leptospirosis. The definition of severe leptospirosis varied in different studies, but for the purposes of this review it included death or patients with a more complicated clinical course. There were 35 articles that satisfied the review’s inclusion criteria. Increasing age was asso-ciated with severe disease in 7 studies. Abnormal respiratory examination findings (18 stud-ies), hypotension (11 studies), oliguria (8 studies), jaundice (7 studies) and altered mental status (4 studies) also helped identify high-risk patients. Abnormal laboratory tests – specifi-cally the complete blood count (17 studies), measures of renal function (16 studies) and liver function (14 studies) – were also associated with severe disease. There was geographical heterogeneity in the clinical phenotype of severe disease, but the presence of hypotension, respiratory or renal involvement had prognostic utility in all regions. Simple bedside findings and basic laboratory tests can provide valuable clinical information in patients with lepto-spirosis. Integration of these indices into early risk stratification tools may facilitate recogni-tion of the high-risk patient and expedite escalation of care in resource-limited settings where most cases of life-threatening leptospirosis are seen.

Abstract: This study will focus on the connection between financial overconfidence and financial vulnerability, using data from the 2024 FINRA National Financial Capability Study (NFCS). The main question was whether people who think they know more about money than they actually do would end up taking more financial risks. After applying Exploratory Factor Analysis, the study identified three key areas that include financial confidence, financial knowledge, and financial vulnerability. These were combined to create an overconfidence score that was then used in several prediction models. Five machine learning models were tested on ten random sample seeds. Among these tested models, XGBoost performed the best to have explained about 60% of the variation in financial vulnerability. To better understand the results, SHAP analysis was then applied to see which factors mattered most. Confidence turned out to be the strongest influence, followed by overconfidence and knowledge. Income and education had smaller effects. Surprisingly, the results did not show that overconfidence increases vulnerability. Instead, a moderate level of confidence actually tends to be financially protective, especially for people with lower knowledge levels. Overall, it was figured that confidence may act more as a form of psychological support than as a risk factor. This finding suggests that financial education should help people build both knowledge and self-belief.

Abstract: Background: A large number of guideline recommendations have been developed by local-, state- and national-level health organisations available to promote the healthy eating and physical activity of children attending early childhood education and care (ECEC) settings. However, the evidence supporting these recommendations is often not well-described. An examination of the current evidence is needed to support decision-makers to understand and prioritise practices for implementation.Aim: To describe a novel systematic evidence-mapping process which: i) examines the evidence-base underpinning ECEC-based healthy eating and physical activity practice elements; and ii) classifies practice elements according to the World Health Organization (WHO) Standards for Healthy Eating, Physical Activity, Sedentary Behaviour and Sleep in Early Childhood and Care Settings to examine alignment with current global guidelines. Methods: We undertook a two-stage, five-step systematic process. Stage 1 involved identifying the existing ECEC-based guideline recommendations and randomised controlled trial evidence which evaluated child diet and physical activity outcomes. Stage 2 involved conducting a secondary data analysis and synthesis of the evidence underpinning practice elements by: extracting practice elements of RCTs and mapping these to existing guideline recommendations, where possible, or included these as additional practices; using vote-counting approaches and a framework to assess the evidence underpinning each practice element; and, classifying practice elements according to the WHO Standards for Healthy Eating, Physical Activity, Sedentary Behaviour and Sleep in Early Childhood and Care Settings. Results: We found 16 healthy eating (e.g. Educators discuss the food served with children) and 19 physical activity (e.g. Educators embed physical activity into educational activities) practice elements were assessed as likely beneficial. Most of these mapped to WHO Standard 2: Creating supportive environments. Seven practice elements were assessed as possibly beneficial, two as possibly not beneficial and none as not beneficial. There was insufficient evidence to assess 39 practice elements. Conclusions: This study provides insights into the evidence underpinning practice elements included in ECEC-based healthy eating and physical activity guidelines, identifies evidence-based practice elements not included in existing guidelines and highlights opportunities where evidence can be strengthened.

Abstract: Uncertainties in geometry, material properties, and excitation forces critically influence the performance of nonlinear electromagnetic vibration energy harvesters, which are promising power sources for wireless sensor networks in industrial environments. These nonlinear harvesters rely on tunable magnetic stiffness to achieve broadband operation, but their strong nonlinear coupling makes them highly sensitive to small parameter deviations. This study investigates how geometric tolerances, variability of magnetic material properties, and excitation irregularities affect the dynamic response and harvested output power of electromagnetic vibration energy harvesters. Nonlinear magnetic restoring forces were obtained using Finite Element Method Magnetics simulations and implemented in a one-degree-of-freedom model for numerical analysis. The results show that deviations as small as ±0.1 mm in geometry or ±5 % in magnetic coercivity can shift the system between monostable, bistable, and chaotic regimes, which could dramatically change wireless sensor operation. Controlled asymmetry of design and impulsive excitation were found to facilitate high-energy orbits, enhancing stability and energy conversion. These findings demonstrate that understanding and managing uncertainty amplification across geometric, material, and excitation domains is essential for reproducible and reliable operation, supporting the design of robust nonlinear electromagnetic harvesters for industrial applications of wireless sensor networks.

Abstract:

Introduction: Chronic obstructive pulmonary disease (COPD) is a progressive condition and a leading cause of morbidity and mortality worldwide. The rate of forced expiratory volume in one second (FEV1) decline is a key prognostic marker. This study aimed to evaluate the impact of demographic, clinical, and therapeutic factors, including respiratory muscle strength and inhaled corticosteroid (ICS) use, on FEV1 decline.Methods: A prospective study was conducted in 2019 at the Clinic for Pulmonology, Clinical Center of Serbia. Fifty patients with stable COPD underwent spirometry, body plethysmography, respiratory muscle strength testing, and laboratory analyses. Demo-graphic and clinical data were collected via questionnaire. All assessments were repeated after six months of regular inhalation therapy.Results: Significant reductions were observed in FEV1, FVC, TLC, PImax, and PEmax. Lower baseline respiratory muscle strength predicted a faster FEV1 decline. Patients in GOLD stage 2 and those with greater hyperinflation exhibited accelerated functional deterioration. Therapy type affected selected parameters: LABA/ICS and LAMA/LABA/ICS regimens significantly reduced residual volume, but not the rate of FEV1 decline. Higher eosinophil counts were associated with a slower reduction in FEV1, suggesting a potential protective effect of ICS.Conclusion: Respiratory muscle strength, hyperinflation, and eosinophil count represent important predictors of COPD progression. Incorporating these parameters into diagnostic and therapeutic algorithms may improve early risk stratification and support indi-vidualized treatment strategies.

Abstract:

The intracellular retention of misfolded extracellular matrix proteins is a common disease mechanism in various individually rare skeletal diseases. This discovery has driven the study of ER-stress and the unfolded protein response (UPR) as a promising therapeutic target in several skeletal dysplasias. In the case of COL10A1 mutations, targeting the UPR resulted in a clinical trial of the repurposed drug carbamazepine; however, for other closely related skeletal disorders treatment with carbamazepine was ineffective, indicating the need for suitable markers for in vitro screenings of potential drug treatments. Mutations in cartilage oligomeric matrix protein (COMP), a cartilage structural protein, causes both multiple epiphyseal dysplasia (MED) and pseudoachondroplasia (PSACH); together referred to as COMPopathies, and result in intracellular retention of mutant COMP protein to varying degrees. In contrast to other closely related skeletal disorders, caused by mutations in cartilage structural proteins, the involvement of the UPR is less clear, and so far, no common COMPopathy marker has been identified. Here, using cell models of COMPopathies we identified MMP9 upregulation as a common feature of six pathogenic COMP variants that do not induce a prominent UPR. We further show that the archetypal p.V194D matrilin-3 MED variant (which causes MED), does not induce MMP9 expression, suggesting that MMP9 upregulation could serve as a specific marker of COMPopathies in vitro.

Abstract: Irritable bowel syndrome (IBS) is a common gastrointestinal disorder marked by abdominal discomfort, altered bowel habits, and comorbidities like anxiety, fatigue, and poor sleep. Existing treatments often fall short of long-term relief and fail to address root causes, leaving many patients with persistent symptoms. This contributes to a substantial burden on individuals and healthcare systems, highlighting the need for new, mechanism-based therapies. Emerging evidence suggests that gastrointestinal symptoms may be caused by the cumulative burden of epithelial injury and insufficient time for mucosal maintenance and repair. This paper explores the hypothesis that structured intermittent fasting regimes could represent a safe, low-cost, and underutilized therapeutic strategy for IBS by promoting gut renewal and restoring homeostasis. We review the multifactorial pathophysiology of IBS and explore how fasting may counter these mechanisms. Evidence from IBS and IBD studies shows that intermittent fasting can reduce inflammation, enhance autophagy, regulate gut motility, and reshape the microbiota, thereby strengthening the gut barrier and dampening immune responses. Notably, fasting induces autophagy, a key cellular recycling process essential for intestinal barrier maintenance and microbial defense, which may be impaired in IBS. Although clinical studies on fasting regimes and IBS are limited, evidence from related populations and mechanistic research supports further exploration. As a practical, circadian-aligned approach that does not restrict specific foods, intermittent fasting may reduce epithelial injury and allow time for repair. The shift from “what to eat” to “when to eat” offers a new, physiology-based potential tool for IBS treatment.

Abstract: Purpose: In order to better understand the relationship between nutrition and brain health, we compared different studies and their results to evaluate the significant association between diet and preventing psychological & neurological disorders such as: Alzheimer’s disease, depression or how different diet patterns can influence mental and brain health. Method: This research includes a descriptive and retrospective study, including a number of 280 subjects of both sexes and different ages, who accomplished a proposed survey. A statistical analysis was made targeting a variety of parameters that indirectly reflects the impact of diet on mental health. Results: General health– 46,8% of subjects confirm the presence of a high impact on general health (131 cases), followed by the category thinking that chose the presence of a very high impact (114 cases, 40,7%). Physical health – in this case the statistical analysis highlighted equal percentages among high (127 cases, 45,4%) and very high impact (126 cases, 45,0%) on physical health. Mental health – 43,6% of subjects consider that adopting a healthy diet is followed by a significant impact mostly on mental health (122 cases, 43,6%). Focus capacity– 43,2% of subjects consider that healthy alimentation has a significant effect on the ability to focus (121 cases). The ability to memorise – similarly to the previous section, the analysis indicated in the previous item that diet has a high impact on the capacity to memorise (109 cases, 38,9%). Good disposition – the major impact is shown is shown on this aspect as well (125 cases, 44,6%). Conclusion: Over 70% of the subjects involved in this research paper have healthy eating behaviours in correlation with the high percentage of patients without any neuro-degenerative or mental health disorders. Most of the subjects consider that healthy nutrition has a significant impact on health in all its forms (general, mental, physic), but not all of them have healthy food habits, nor use their knowledge. Promising results show that healthy nutrition positively contributes in alleviating focus or learning capacity. There has been a distinctive link between daily caffeine consumption which is associated with an unhealthy diet and unhealthy diet behaviours. 24,6% of subjects with an unhealthy diet and 4,9% of those with relative healthy diet agree daily consumption of soda drinks. Negative feelings, with a significant emotional impact were reported in most of the cases or in superior proportions in subjects with an unhealthy diet (prevailing or intermittent), proving that unhealthy habits might have an important effect on mental health, while exacerbating mental health disorders.

Abstract: Piggery farming is the largest source of livestock manure in South Korea, generating about 40% of total livestock waste annually. Yet greenhouse gas (GHG) emission data from piggery wastewater treatment systems remain limited, with most studies focused on farm slurry storage rather than process-level emissions. This study quantified methane (CH₄) and nitrous oxide (N₂O) fluxes from a full-scale piggery wastewater treatment facility in, to develop process, season, specific and diurnal specific emission fluxes. Continuous monitoring with a laser-based gas analyzer and customized PVC air-pool chamber was conducted across raw, anaerobic, and aerobic wastewater treatment stages. Mean CH₄ fluxes ranged 1.1-15.6 mg s⁻¹ m⁻², peaking in summer, while N₂O fluxes ranged 0.01-17971 mg s⁻¹ m⁻², with maxima in fall. Aeration tank II and Anaerobic tank I were the dominant emission stages, with night and intra-day peaks. Statistical analysis identified treatment stage and temperature as the main controls on CH₄ variability (p = 0.006 to 0.014), whereas N₂O showed weaker climatic sensitivity. The results provide refined emission factors and emphasize that aeration optimization and denitrification control are key to reducing GHG emissions from livestock wastewater systems in warm, humid regions.

Abstract: We present a systematic evaluation of large language models on quantum mechanics problem-solving. Our study evaluates 15 models from five providers (OpenAI, Anthropic, Google, Alibaba, DeepSeek) spanning three capability tiers on 20 tasks covering derivations, creative problems, non-standard concepts, and numerical computation, comprising 900 baseline and 75 tool-augmented assessments. Results reveal clear tier stratification: flagship models achieve 81% average accuracy, outperforming mid-tier (77%) and fast models (67%) by 4pp and 14pp respectively. Task difficulty patterns emerge distinctly: derivations show highest performance (92% average, 100% for flagship models), while numerical computation remains most challenging (42%). Tool augmentation on numerical tasks yields task-dependent effects: modest overall improvement (+4.4pp) at 3x token cost masks dramatic heterogeneity ranging from +29pp gains to -16pp degradation. Reproducibility analysis across three runs quantifies 6.3pp average variance, with flagshipmodels demonstrating exceptional stability (GPT-5 achieves zero variance) while specialized models require multi-run evaluation. This work contributes: (i) a benchmark for quantum mechanics with automatic verification, (ii) systematic evaluation quantifying tier-based performance hierarchies, (iii) empirical analysis of tool augmentation trade-offs, and (iv) reproducibility characterization. All tasks, verifiers, and results are publicly released.

Abstract: Seed aging during storage represents a major challenge to global food security and germplasm resource conservation. Long-lived mRNAs, which are crucial for initiating germination after storage, have poorly understood regulatory mechanisms governing their stability. In this study, we identify the RNA-binding protein OsGRP3 as a key positive regulator of rice storability. Initially, we demonstrated that Arabidopsis AtGRP7 enhances seed vigor following aging. Phylogenetic analysis identified OsGRP3 as its closest rice homolog. Two independent OsGRP3-overexpression lines showed markedly improved germination rates and seed viability after extended artificial aging. Physiological assessments indicated that OsGRP3 mitigates aging-related damage, as evidenced by reduced malondialdehyde (MDA) levels and electrolyte leakage, consistent with better membrane integrity. RNA-seq analysis revealed that OsGRP3 overexpression attenuated the transcriptional disruption induced by aging. Moreover, OsGRP3 activates a "pre-adaptive" transcriptional network under non-stress conditions, leading to altered expression of 404 genes involved in DNA replication, gluconeogenesis, and essential amino acid metabolism to the onset of aging stress. Our findings establish OsGRP3 as a conserved RNA-binding protein that enhances seed storability, and offer a promising genetic target for improving storage tolerance in rice.

Abstract: Classical macroscopic models of metal–hydride (MH) hydrogen storage rely on empirical Arrhenius laws that neglect quantum phenomena such as tunneling, zero-point motion, and hydrogen–lattice interactions. As a result, their predictive and control performance degrade across wide temperature ranges, particularly in cryogenic regimes where quantum transport remains active. This paper presents a unified quantum-informed diffusion and control framework that bridges microscopic hydrogen–lattice physics with macroscopic predictive control. A temperature-dependent quantum correction operator is incorporated into the classical diffusion law, yielding an analytically tractable yet physically enriched model. Parameters are identified through weighted robust regression with bootstrap-based uncertainty quantification and integrated into a model predictive control (MPC) scheme that adapts to temperaturedependent dynamics. Simulation results show that tunneling-enhanced diffusion improves lowtemperature response and reduces steady-state error and control effort by up to 50% compared with classical Arrhenius-based control. While the present study focuses on numerical validation, the proposed architecture establishes a transferable foundation for digital-twin development—linking microscopic quantum transport and system-level predictive control for next-generation hydrogen storage technologies.

Abstract: Diffuse large B-cell lymphoma (DLBCL) is the most common and clinically aggressive subtype of non-Hodgkin lymphoma (NHL). Although novel therapeutic agents, in-cluding rituximab and polatuzumab vedotin, have improved outcomes, almost one-third of patients ultimately develop relapsed or refractory disease. MicroRNAs (miRNAs), a class of endogenous single-stranded RNAs approximately 22 nucleotides in length, play a pivotal role in the regulation of gene expression at the post-transcriptional level through interactions with complementary target RNAs and contribute significantly to the de-velopment, progression, and treatment response of DLBCL. Oncogenic miRNAs, such as miR-155, miR-21, and the miR-17–92 cluster, promote proliferation, survival, immune evasion, and therapy resistance by modulating pathways including PI3K/AKT, NF-κB, and MYC. Conversely, tumor-suppressive miRNAs such as miR-34a, miR-144, miR-181a, and miR-124-3p inhibit oncogene activity and enhance apoptosis, with their loss often associated with adverse outcomes. Among these, miR-155 and miR-21 are particularly well studied, playing central roles in both tumor progression and remodeling of the tumor microenvironment. This review summarizes current evidence on the biological and clinical relevance of miRNAs in DLBCL, emphasizing their diagnostic and prognostic potential.

Abstract: This article critically examines the role of financial education as a catalyst for financial freedom, understood not only as economic autonomy, but also as the real capacity for agency and deliberation about material life. Based on a qualitative, descriptive, and documentary study, a thematic content analysis was applied to academic literature, institutional reports, and classic and contemporary works on personal finance.The results show that financial literacy, although necessary, is insufficient if it is not articulated with psychological, socio-emotional, and cultural dimensions, such as financial mindset, discipline, emotions, and beliefs about money. From a transdisciplinary approach, the interrelationships between financial education, self-education, the psychology of money, and decision-making are explored, proposing a comprehensive model that links knowledge, action, and being.It is argued that financial education must transcend the mere transmission of technical content to become a process of ethical and political empowerment, oriented toward the development of sustainable habits, critical thinking, and socio-emotional skills that allow for conscious control over the economy and thoughtful participation in the structures that shape collective life.It concludes that, in order to be transformative, financial education must be multidimensional, situated, and adaptive, capable of responding to the challenges of a global environment characterized by inequality and technological acceleration. It proposes moving toward evidence-based programs that strengthen civic agency and deliberative participation as constitutive dimensions of contemporary political subjectivity.

Abstract: Botanical gardens in Ethiopia function as vital socio-ecological systems supporting biodiversity conservation, cultural heritage, environmental education, and climate resilience. This study conducts a multi-dimensional evaluation of three major botanical gardens Gullele (GUBG), Shashemene (SHBG), and Dilla University (DUBEG) using mixed methods involving 300 stakeholder surveys, 15 interviews, and field observations. Six performance domains were assessed: governance, research, education, infrastructure, health and well-being, and cultural integration. Quantitative results indicate that Gullele achieved the highest performance score (mean 4.08), attributed to effective governance and strong infrastructure. Shashemene performed best in cultural integration, while Dilla University excelled in research. Logistic regression highlighted governance and infrastructure as key predictors of institutional success. Qualitative analysis revealed persistent challenges, including fragmented mandates, unstable funding, low community participation, and infrastructural deficits limiting long-term sustainability. Despite these barriers, Ethiopian botanical gardens show substantial potential to advance the nation’s Climate-Resilient Green Economy and Sustainable Development Goals. Strengthening coordinated governance, diversifying funding sources, and promoting local knowledge systems are essential for improving institutional resilience. Enhancing these gardens’ capacities will reinforce their contributions to sustainable land management, biodiversity protection, climate adaptation, and public well-being within Ethiopia’s diverse ecological and cultural landscapes.

Abstract: Water Alternating Gas (WAG) injection is a well-established enhanced oil recovery technique that improves sweep efficiency by combining the favorable displacement characteristics of waterflooding and gas injection. This work presents a sequential Picard–Newton formulation for simulating three-phase flow under WAG conditions in heterogeneous petroleum reservoirs. The mathematical model considers slightly compressible, immiscible oil, water, and gas phases under isothermal conditions, discretized using the finite volume method. Reservoir heterogeneity is represented through geostatistical permeability fields generated by Sequential Gaussian Simulation, capturing the spatial correlations and anisotropy characteristic of subsurface formations. The methodology is applied to investigate WAG performance in two heterogeneous reservoir models with mean permeabilities of 100 mD and 200 mD under identical 1:1 injection ratios. The numerical results successfully reproduce the cyclic saturation and production behavior characteristic of WAG processes. Comparative analysis reveals that higher permeability enhances injectivity and cumulative recovery but accelerates water breakthrough and production decline, illustrating the trade-off between displacement efficiency and sweep control. These findings demonstrate that the proposed framework provides an efficient and physically consistent tool for evaluating WAG strategies in heterogeneous reservoirs, with potential application to field-scale optimization of advanced recovery operations.

Abstract: Background/Objectives: Next to malaria, respiratory viruses, particularly respiratory syncytial virus (RSV), are responsible for the hospitalization and death of thousands of young children each year in sub-Saharan Africa. During peak seasons, conducting separate tests is time-consuming and distressing. This underscores the need for efficient, rapid multiplexed diagnostic tools. This study evaluated the clinical performance of a lateral flow assay (LFA) based antigen combo rapid diagnostic test (ML Ag Combo RDT, manufactured by MobiLab) that detects RSV, influenza A and B, and SARS-COV-2. Methods: The Allplex rRT-qPCR assay was used as a reference assay to evaluate the clinical performance of the LFA Ag Combo RDT in pediatric hospital settings and performed using 470 nasopharyngeal swab (NPS) specimens from hospitalized children under two with respiratory symptoms. Results: Based on the comparative analysis of the testing results for 470 NPS, the ML Ag Combo RDT demonstrated high sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 90.06%, 98.33%, 93.45%, and 97.38% for RSV, and 71.43%, 100%, 100%, and 92.94% for influenza A/B, respectively. Agreement with the Allplex rRT-qPCR was strong (κ = 0.90 for RSV and 0.82 for Flu A/B), with overall accuracies of 96.60% for RSV and 94.47% for Flu A/B. This was further supported by ROC analysis for aggregated data (RSV and Flu) with an AUC value of 0.925. As expected, in samples with high viral loads (Ct < 20), the Ag Combo RDT achieved 100% sensitivity for RSV and Flu A/B. Sensitivity declined slightly at lower viral loads (Ct > 35). Conclusions: The ML Ag Combo RDT demonstrates high specificity and strong diagnostic accuracy for the detection of RSV and Flu A/B in pediatric hospital settings, where timely diagnosis is critical.

Abstract: Background: Rheumatoid arthritis (RA) is a chronic inflammatory condition affecting joints. It is managed with pharmacological therapies, but it is speculated that dietary intervention may be of significant benefit to patients in helping to manage the condition alongside their medication. This may be mediated through effects on levels of chronic inflammation and thus can influence disease activity. Previous research has indicated the potential of diet to aid in the condition’s management, but a comprehensive review is required to pull together the evidence. Objective: Conduct a systematic review of existing literature on the impact of differing dietary interventions on disease activity and inflammatory status in rheumatoid arthritis patients. Then draw conclusions about the strength of evidence and identify patterns amongst identified papers in a narrative synthesis. Methods: The existing literature was searched according to a pre-determined search strategy, using the databases PubMed, Embase and MEDLINE. Data was then extracted according to a template and results related in a narrative synthesis. Meta-analysis was not conducted. Results: Eight studies, all randomised controlled trials, were included in the review. Outcomes included disease activity score 28 and inflammatory markers such as erythrocyte sedimentation rate and C-reactive protein. Dietary patterns such as an anti-inflammatory diet showed inconclusive evidence with regards to impact on disease outcomes. Strong evidence for the reduction of inflammation and disease activity was demonstrated in studies investigating probiotic supplementation. Conclusion: Results demonstrated the ability of probiotics to have a beneficial impact for RA patients, which may allow healthcare professionals to develop treatment plans and advice including this intervention. However more research should be conducted in different regions and in combination with other interventions to further the understanding of its effects and safety. Other dietary interventions included showed significant associations for some outcomes and not others. Therefore, further research should be conducted to determine whether they are feasible for use in the management of RA patients.

Abstract: Accurate medical image segmentation plays a crucial role in clinical diagnosis and treatment planning. Yet, existing U-Net variants often fail to effectively capture both fine anatomical details and global contextual relationships while maintaining efficiency. To overcome these challenges, we propose EB-MSF-UNet, an Efficient Boundary-aware Multi-Scale Feature Fusion U-Net that integrates global reasoning with local detail preservation. The framework employs a dual-pathway encoder combining lightweight convolutional layers for local feature extraction and Mamba-inspired modules for longrange dependency modeling, regulated by an adaptive feature gating mechanism. A cross-attentionbased multi-scale fusion module ensures consistent interaction across feature hierarchies, while a boundary-aware refinement decoder explicitly enhances structural contours through auxiliary boundary supervision. Experiments on multiple medical image benchmarks demonstrate that the proposed method achieves consistently accurate segmentation with improved boundary precision and computational efficiency.

Abstract: A cascade deep-learning approach is proposed for optimizing the design and control of a dual-frequency induction heating system used in semiconductor manufacturing. The system is composed of two independent power inductors, fed at different frequencies, to achieve a homogeneous temperature profile along a graphite susceptor surface, crucial for enhancing layer quality and integrity. The optimization process considers both electrical (current magnitudes and frequencies) and geometrical parameters of the coils, which influence the power penetration and subsequent temperature distribution within the graphite disk. A two-step procedure based on Deep Neural Networks (DNNs) is employed. The first step, namely optimal design, identifies the optimal operating frequencies and geometrical parameters of the two coils. The second step, namely optimal control, determines the optimal current magnitudes. The DNNs are trained using a database generated through Finite Element (FE) analysis. The optimized system successfully achieved significant thermal uniformity across the susceptor surface, such that the temperature fluctuations over the entire disk radius were kept within ±7% of the target temperature (1100 °C). This significant improvement confirms the effectiveness of the deep learning-based cascade approach for solving high-dimensional Multiphysics inverse problems.