26734 research outputs found
Sort by
PRagMatic Pediatric Trial of Balanced vs nOrmaL Saline FlUid in Sepsis: study protocol for the PRoMPT BOLUS randomized interventional trial
No full textAbstract Background/aims Despite evidence that preferential use of balanced/buffered fluids may improve outcomes compared with chloride-rich 0.9% saline, saline remains the most commonly used fluid for children with septic shock. We aim to determine if resuscitation with balanced/buffered fluids as part of usual care will improve outcomes, in part through reduced kidney injury and without an increase in adverse effects, compared to 0.9% saline for children with septic shock. Methods The Pragmatic Pediatric Trial of Balanced versus Normal Saline Fluid in Sepsis (PRoMPT BOLUS) study is an international, open-label pragmatic interventional trial being conducted at > 40 sites in the USA, Canada, and Australia/New Zealand starting on August 25, 2020, and continuing for 5 years. Children > 6 months to < 18 years treated for suspected septic shock with abnormal perfusion in an emergency department will be randomized to receive either balanced/buffered crystalloids (intervention) or 0.9% saline (control) for initial resuscitation and maintenance fluids for up to 48 h. Eligible patients are enrolled and randomized using serially numbered, opaque envelopes concurrent with clinical care. Given the life-threatening nature of septic shock and narrow therapeutic window to start fluid resuscitation, patients may be enrolled under “exception from informed consent” in the USA or “deferred consent” in Canada and Australia/New Zealand. Other than fluid type, all decisions about timing, volume, and rate of fluid administration remain at the discretion of the treating clinicians. For pragmatic reasons, clinicians will not be blinded to study fluid type. Anticipated enrollment is 8800 patients. The primary outcome will be major adverse kidney events within 30 days (MAKE30), a composite of death, renal replacement therapy, and persistent kidney dysfunction. Additional effectiveness, safety, and biologic outcomes will also be analyzed. Discussion PRoMPT BOLUS will provide high-quality evidence for the comparative effectiveness of buffered/balanced crystalloids versus 0.9% saline for the initial fluid management of children with suspected septic shock in emergency settings. Trial registration PRoMPT BOLUS was first registered at ClinicalTrials.gov ( NCT04102371 ) on September 25, 2019. Enrollment started on August 25, 2020
Early Permian Stratigraphy and Sedimentology, NW Devon and SW Ellesmere Islands, Sverdrup Basin (Arctic Canada, Nunavut)
No full textThis study investigates the Upper Pennsylvanian (Gzhelian) to Lower Permian (Asselian– Artinskian) succession on NW Devon Island (Grinnell Peninsula) and SW Ellesmere Island (Bjorne Peninsula) along the southern margin of the Sverdrup Basin. Integrated facies and stratigraphic analyses delineate four third-order stratigraphic sequences of Gzhelian, early to middle Asselian, late Asselian–late Sakmarian, and latest Sakmarian–late Artinskian age. These record a complex interplay of sedimentation and episodic tectonic activity, with differential uplift and subsidence influencing facies distribution and thickness. Warm water, photozoan assemblages existed in the Gzhelian, but were overwhelmed by significant clastic influxes that diminished into the Early Permian. Gzhelian-aged deposits on Grinnell Peninsula are notably thin, a feather edge of considerably thicker deposits elsewhere in the basin. Glacio-eustatic cyclothems within the Asselian Belcher Channel Formation led to strata with characteristic resistant- recessive bands within a prolific photozoan carbonate factory, most notably Palaeoaplysina buildups. This high carbonate productivity ceased in the latest Asselian, primarily driven by the closure of the Uralian seaway which triggered a transition from warm photozoan to cool heterozoan carbonate factories into the Sakmarian. A narrow band of heterozoan-extended carbonates developed at the periphery of the basin indicating warm seawater temperatures were maintained in the shallowest of waters, that rapidly transition to heterozoan mudrock deposits within the rapidly subsiding area around Bjorne Peninsula. By the Artinskian, carbonate production was exclusively heterozoan, reflecting cooler oceanic conditions across the Sverdrup Basin
‘We truly feel limited’: nurses and midwives’ perspectives on multi-level factors influencing women’s adherence to ANC in Rwanda: a qualitative study
No full textAbstract Background Antenatal care (ANC) is essential for improving maternal and neonatal health outcomes, yet adherence to ANC services remains a challenge in many low-income settings, including Rwanda. Understanding nurses and midwives’ perspectives on factors influencing ANC adherence is crucial for developing targeted interventions to enhance service utilization and maternal health outcomes. Methods This study employed a qualitative descriptive design to explore the perspectives of nurses and midwives on ANC adherence in Rwanda. Fifteen in-depth interviews (IDIs) were conducted using a semi-structured interview guide in Kinyarwanda. The interviews were verbatim transcribed and then translated into English. Atlas.ti 7 software was used to organise the data and then thematically analysed. Results The perspectives of nurses and midwives were summarised in four themes. Participants mentioned facilitators of ANC engagement with ANC services such as community education, structural motivators, availability of diagnostic infrastructure like ultrasound, and nurses and midwives training and mentorship. The barriers to women’s ANC adherence noted by participants are cultural beliefs and community misconceptions, stigma and secrecy surrounding unintended pregnancies, cost-related delays in ANC seeking, gender dynamics and relationships. Nurses and midwives also highlighted health care system constraints, such as staffing shortages and infrastructure and equipment limitations. Recommended interventions to enhance ANC adherence included community engagement and support, increased staff and resources, and digitalization of records. Conclusion Nurses and midwives play a critical role in shaping ANC adherence through service delivery and patient education. Their consistent engagement and ability to build trust with pregnant women make them key influencers in promoting timely and sustained ANC attendance. Addressing systemic challenges, strengthening community-based support, and enhancing policy implementation are essential strategies for improving ANC adherence in Rwanda. These findings provide valuable insights for policymakers and healthcare stakeholders to develop targeted interventions aimed at increasing ANC adherence and improving maternal and neonatal health outcomes
Federated Medical Image Segmentation with Pseudo-Labeling under Model Heterogeneity
No full textMedical image segmentation is fundamental to computer-aided diagnosis and treatment planning, yet its clinical adoption remains constrained by the scarcity of expert-annotated datasets and the inability to centralize patient data due to strict privacy regulations. Manual segmentation is time-consuming and prone to inter- and intra-observer variability, and deep learning models trained on limited, single-institution datasets often fail to generalize across diverse clinical environments. Institutional heterogeneity arising from differences in imaging modalities, scanner protocols, and computational resources further limits the effectiveness of conventional centralized and federated learning approaches. Federated Learning (FL) enables multi-institutional collaboration without raw data sharing, but traditional FL depends on parameter exchange and assumes homogeneous model architectures. These constraints introduce privacy risks, restrict participation from resource-limited institutions, and often lead to unstable performance under non-IID data distributions. To address these limita-tions, we propose MDH-FL, a data-centric, semi-supervised FL framework that eliminates param-eter sharing and supports heterogeneous segmentation model setup tailored to local computational and data characteristics. Instead of exchanging model weights, MDH-FL operates by having clients generate predic-tions and uncertainty scores on a shared unlabeled public dataset. The server applies adaptive pseudo-label aggregation to obtain high-quality pseudo-labels. A two-stage filtering mechanism— i.e., uncertainty-based filtering followed by SSL-guided similarity filtering—selects personalized pseudo-labeled subsets aligned with each client’s data distribution. To simulate real-world hetero- geneity, three U-Net–based architectures with varying complexity were developed across clients. These models employ Exponential Moving Average (EMA) based smoothing to stabilize local updates and ensure consistent learning across federated rounds. The framework is evaluated across two imaging modalities: Breast Cancer Ultrasound (BCU), using three heterogeneous ultrasound datasets, and Skin Cancer Dermoscopy (SCD), using five dermoscopy datasets. Experiments demonstrate that MDH-FL achieves 2–9% improvements over client-specific non-FL baselines and 4–20% gains over conventional FL methods such as FedAvg, FedNova, and FedProx. These results highlight MDH-FL’s robustness, stability, and capacity to support heterogeneous model architectures without compromising privacy. Overall, MDH-FL provides a scalable, privacy-preserving, and architecture-agnostic frame-work for generalizable medical image segmentation, offering a practical pathway for real-world deployment in diverse and resource-constrained healthcare environments
Performance Improvement in Robotic Milling: Hybrid Compliance Error Compensation and Fast-Chirp Identification Approach
No full textIndustrial robots have been used in machining over the past two decades due to advantages such as lower cost and a greater ability to machine large parts and sculpted geometries compared to CNC machine tools. However, their low stiffness causes deflection and vibration during high-force tasks, which limits their application in machining. This study aims to address these challenges by developing a dynamic model of industrial robots. A dynamic model incorporating joint tri-axial flexibility is developed. The robot’s rigid-body model is identified, and a new friction model is proposed to reduce the identification error. The limitations of the physics-based model, with a focus on friction nonlinearities, are investigated. To further reduce the identification error and accurately predict joint torques, a new hybrid torque estimator is introduced. The estimator includes a data-driven scheme based on a recursive linear regression method that uses the output of the identified model along with the joint motion data. This hybrid approach reduces the joints torque estimation error by up to 55 percent. The torque estimator is incorporated into a nonlinear disturbance observer to predict the external torques applied to the joints due to cutting forces in real time, without using any external sensors. The compliance error is calculated from the estimated external torques and the identified joint stiffness, and compensated in real time using a compliance error compensation method. Multi-axis aluminum milling experiments show that the average TCP deflection decreases by 78 percent in the x direction and 77 percent in the y direction. Another contribution of this study is introducing a novel automated approach, termed “fast chirp”, for identifying the joints dynamic parameters, specifically stiffness and damping, in moving robots. A chirp centrifugal force generated by a customized offset mass tool is utilized for two-directional excitation of the robot, providing full control over the excitation within a desired frequency range, and eliminating the need for manual tests. To capture the moving-joints frictional behavior, the robot moves continuously along an identification trajectory which can cover a large portion of workspace in a single experiment, while excitation is applied in multiple zones. A new formulation is developed to calculate vibration responses under simultaneous two-directional excitation, using direct and cross FRFs. This formulation is experimentally validated, and static and quasi-static conditions are compared. The importance of cross-coupling terms and joints tri-axial flexibility assumption in industrial robots is also investigated. The dynamic model and measured vibration responses are used to identify joint stiffness and damping parameters. The identified model is validated and subsequently employed to generate chatter stability lobe diagrams. The accuracy of predicted SLDs is validated through milling chatter experiments. The effect of incorporating the indirect vibration response in the local z direction, which cannot be directly excited using the fast chirp approach, on predicting z-direction dynamics is formulated. Experimental results show that including the z-direction response in the fast chirp identification approach improves the accuracy of the identified model in predicting z-direction dynamics. The identified model is used to predict and avoid chatter in machining applications
Learning to Price; Learning to Hedge: Modernizing Option Pricing and Hedging
No full textThis thesis demonstrates the essential role of machine learning in modernizing option pricing and hedging in the rapidly evolving financial markets. Option pricing and hedging are a very complex but important facet of the financial world. There have been several important studies in literature detailing how to price options, and hedge against future liabilities, using mathematical and numerical techniques. We investigate these two fundamental tasks through distinct learning frameworks: supervised learning for pricing and (in addition) reinforcement learning for hedging. The recent advances of artificial neural networks have proven capable of dealing with today’s complex problems with data. By decoupling valuation from risk man-agement, we highlight how different objectives naturally align with different methodologies. Empirical results on synthetic and market data illustrate that learning-based approaches do not only replicate, and at times, improve on classical results under conventional models, but also adapt to environments where traditional assumptions fail. This work emphasizes the complementary power of data-driven modeling and sequential decision-making in mathe-matical and financial modeling. We specifically investigate the applications of recurrent and convolutional neural networks to help improve the existing solutions to the complex problem of option pricing and hedging
A linked administrative data study of continuity of care for patients involved with Alberta provincial corrections
No full textBackground. Alberta has been a leader in correctional healthcare delivery, providing health services through the Ministry of Health since 2010, and more recently through the Ministry of Mental Health and Addiction. People with incarceration experience tend to have complex health needs spanning mental health, substance use, and physical health. Though continuity has recently become a focus in correctional healthcare, details of patients’ health needs, available services, and continuity of care on release are not widely available outside of Correctional Health Services. Objectives. 1) Characterize people receiving physician care in Alberta corrections between April 1, 2017 and March 31, 2020. 2) Describe fee-for-service healthcare billings from Alberta provincial corrections. 3) Identify factors associated with continuity and fragmentation of care. Methods. Using administrative health data, we identified patients receiving healthcare in Alberta provincial remand and correctional facilities. Multiple logistic regression was used to model predisposing, enabling, need, and behaviour factors associated with adequate continuity and high fragmentation of care. Results were interpreted in collaboration with community partners representing various municipal and provincial organizations. Results. 8,301 patients were identified. Linkage to a primary care physician was the top predictor of adequate continuity and low fragmentation of care. Rural patients were more likely to have adequate continuity. Homelessness and OAT were related to low continuity and high fragmentation. Physical health multimorbidity was related to adequate continuity among patients linked to primary care, and low continuity otherwise. Implications. Administrative data have potential to inform improvements to the care continuum if interpreted with nuance and context. As correctional healthcare shifts its focus to release planning, further work on continuity of care for this population is warranted to improve outcomes for patients upon release
Regional dynamics of zoonotic risk perception and wildlife use in Ghana
No full textAbstract Background The COVID-19 pandemic and other recent zoonotic outbreaks have renewed global scrutiny of the bushmeat trade, particularly in West Africa. This study examines regional differences in bushmeat consumption, hunting practices, and perceptions of zoonotic disease risk in Ghana, using a mixed-methods approach combining a household survey (n = 335) and key informant interviews with hunters (n = 53). Methods Respondents were drawn from northern and southern Ghana to reflect ecological and cultural diversity. Socio-demographic characteristics, bushmeat consumption patterns, and knowledge of zoonotic diseases were analyzed. Hunters were interviewed to explore occupational activities and risk mitigation behaviors. Results Bushmeat consumption declined after the COVID-19 outbreak, more so in southern Ghana (from 62% to 33%) than in the north (from 81% to 61%). Awareness of zoonotic disease transmission was high (~ 70%) and primarily acquired through mass media, yet this had limited impact on behavior. Hunting activity declined during the pandemic, but hunters attributed this not to health concerns, but to wildlife scarcity and reduced commercial demand. Use of personal protective equipment was absent, with hunters citing discomfort, cultural beliefs, and spiritual protection as justifications. Risk perceptions varied regionally: northern respondents emphasized improper cooking as the main transmission pathway, while southerners pointed to handling of live animals. Religious beliefs significantly shaped attitudes toward disease vulnerability, with 80% of southern respondents and 58% in the north attributing protection from disease to divine intervention. Conclusion Bushmeat-related behaviors in Ghana are influenced more by cultural norms, economic necessity, and ecological conditions than by knowledge of zoonotic disease risks. Public health messaging alone is insufficient. Effective interventions must be culturally responsive, integrate conservation with health surveillance, and align with local worldviews. A multidimensional “One Health” approach is essential for sustainable behavior change and zoonotic disease prevention
Estimating E/I Balance from Resting-State fMRI Using Biologically Inspired LSTM Networks: Application to ASD Subtyping
No full textThis thesis investigates Excitatory/Inhibitory (E/I) balance abnormalities in Autism Spectrum Disorder (ASD) through a novel integration of biologically informed Long Short-Term Memory (LSTM) networks and intra-regional connectivity features. Using resting-state functional Magnetic Resonance Imaging (rs-fMRI) data from the Autism Brain Imaging Data Exchange (ABIDE) dataset, a modified LSTM network was developed to simulate the dynamics of excitatory and inhibitory neuron populations. Instead of modeling interactions between regions, the network was independently fitted to the Blood Oxygen Level Dependent (BOLD) time series of each of the 116 brain regions. From this, biologically grounded intra-area E/I connectivity matrices were extracted, capturing local interaction patterns between excitatory and inhibitory units. Additionally, Ne/Ni were computed for each brain region and statistically evaluated across subjects. These ratios, along with intra-area E/I connectivity features, served as input to a Self- Organizing Map (SOM) clustering pipeline, which revealed ASD subtypes characterized by distinct Ne/Ni profiles and associated functional networks. The findings underscore the importance of regional E/I balance, particularly the Ne/Ni, as a key feature in explaining ASD heterogeneity and offer a biologically interpretable framework for identifying subtype-specific patterns relevant to future diagnostics and interventions
Multi-Scale Study of Fluid Flow and Heat Transfer
No full textAn examination of the coupled mechanisms of mass and heat transfer in porous media is presented in this study, ranging from micro-scale physical phenomena to macro-scale applications related to geothermal energy. Research aims to clarify the interrelationship between thermal and solute dispersion and to develop computational and experimental methods for predicting transport behavior in the presence of varying flow conditions. Analytical modeling and numerical simulations were used to establish a consistent correlation between heat and mass dispersion coefficients, demonstrating that dispersion mechanisms in heterogeneous porous structures exhibit similar trends over a wide range of Péclet numbers. As a result of this insight, transport analysis can be simplified by predicting the behavior of one process from the behavior of another. An innovative method has been developed to determine thermal conductivity in both the conductive and convective regimes in order to quantify thermal behavior further. Experimental measurements in a sand-packed porous medium revealed that effective thermal conductivity can increase by several orders of magnitude with fluid velocity, particularly under forced convection. Early-stage temperature peaks, which is attributable to solid-fluid thermal resistance, were observed and identified as critical for accurately capturing transient thermal dynamics. In order to address large-scale applications, a semi-dimensionless modeling approach was developed to predict temperature changes along a geothermal wellbore. With this method, the governing equations are transformed into a dimensionless form in the longitudinal direction, resulting in a significant reduction in the computational cost while maintaining high accuracy. Based on simulation results validated against experimental data, it was determined that injection strategies, insulation techniques, and well geometry are significant factors influencing heat recovery efficiency. When annulus injection is conducted with the appropriate insulation, higher outlet temperatures are consistently achieved, especially in deeper wells and at lower injection rates. The findings contribute to a multi-scale framework for understanding and optimizing heat and mass transport in porous media. It advances predictive capabilities for energy extraction systems by integrating pore-scale mechanisms with laboratory experimentation and computational modeling, while providing practical tools for improving the design and operation of geothermal and subsurface thermal technologies