136986 research outputs found
Sort by
Adaptation of Connection Systems for Integration with Engineered Wood Products in Buildings: A Systematic Review
No full textConnection systems are a critical component of buildings constructed with engineered wood products (EWPs), influencing structural integrity, durability, and construction efficiency. This systematic review categorises connection types into mechanical, adhesive, and interlocking systems and evaluates their structural performance, adaptability in prefabrication, applicable design standards, and modelling approaches. The review synthesises recent trends in EWP connection research, highlighting key developments in digital fabrication, reversible joints, and sustainable construction. Findings emphasise the need for standardisation, performance validation, and hybrid systems to support the wider adoption of prefabricated timber structures in environmentally responsible building practices
Automated and Enhanced Choice Modelling: Estimation and Specification
No full text© 2025 Amir GhorbaniDiscrete choice models (DCMs) are essential tools in transportation planning, offering insights into travel behaviour, demand prediction, and the evaluation of policies and projects. These models help planners understand how individuals make decisions regarding travel modes, destinations, paths and activities based on factors such as travel time, cost, and service quality. By simulating potential changes in choice attributes, DCMs enable predictions of future demand under various scenarios, assisting in the assessment of infrastructure investments, pricing strategies, and policy interventions that affect network performance and sustainability.
However, the specification of DCMs involves a complex process that requires critical decisions about which variables to include, their functional forms, required transformations, and interactions with other factors, and the assumptions about distributional properties for the coefficients and error terms. A challenge in this field is achieving both predictability and interpretability. Predictability refers to a model's ability to forecast choice labels, either deterministically or probabilistically. Interpretability, on the other hand, is twofold: it includes explainability, which refers to the ability to clarify how input attributes influence outcomes (such as through marginal effects or willingness to pay), and trustworthiness, which ensures that the model adheres to domain-specific knowledge and constraints that experts may consider 'a priori', such as a monotonic decrease in utility with respect to cost or expected higher values for travel time savings for specific modes of transport. Relying solely on a choice model’s predictive performance is insufficient to justify its use. For instance, one might obtain several models - both hand-crafted and data-driven - with comparable predictive accuracy on a given dataset, yet the trustworthiness of their interpretability - for example, the derived willingness-to-pay estimates - can differ markedly. Which model, then, should we trust?
Achieving both predictability and trustworthiness in DCMs presents several challenges. Traditional DCMs often rely on manually crafted utility specifications, which could be more interpretable than data-driven models but require significant input from the modeller to avoid misspecification and ensure trustworthiness. However, this approach can suffer from limited predictive accuracy, especially when dealing with large datasets that involve numerous attributes, and it is prone to oversimplification of human behaviour. To make the process more efficient, optimisation-based algorithms have been proposed to search for the best utility functions. Additionally, data-driven approaches have been introduced, such as neural networks and kernel-based methods, which enhance predictive power by capturing complex attribute interactions.
Despite these advancements, there are still gaps in ensuring trustworthiness in data-driven models. Previous research has made attempts to integrate monotonicity constraints into data-driven models to preserve some aspects of behavioural consistency, but more comprehensive methods are needed to incorporate broader behavioural insights while maintaining predictive accuracy.
In my research, I have tackled these challenges with three primary contributions. First, I addressed predictability by developing a new estimation technique called DUET: Deterministic-probabilistic Utility Estimation Technique that improves the core estimation approach, which has traditionally relied solely on maximum likelihood estimation (only probabilistic). Our results reveal a 6.5 per cent increase in predictability compared to the state-of-the-art deep neural network model. Second, I proposed a modelling framework called BINN: Behaviourally-Informed Neural Network to address behavioural trustworthiness and automate model specification by embedding domain-specific constraints directly into the neural network architecture. This framework ensures that the model not only predicts accurately but also aligns with expert expectations, such as monotonicity in utility functions. Both of these contributions were evaluated using benchmark mode choice data from the Swissmetro dataset as well as synthetically generated datasets, where they showed improvements in both model performance and trustworthiness.
Lastly, I introduce Infinity, a conversational artificial intelligence assistant for choice-modelling specification. Such a tool is particularly valuable for data-driven choice-modelling approaches, which are often harder to tailor to specific contexts and require specialist knowledge to adapt effectively. Infinity represents the first AI assistant designed to support data-driven choice models, which is trained to address user needs in applying the proposed BINN model.
It is worth noting that all three contributions of this thesis advance the automation of choice model specification; each chapter explains the individual contribution in detail. In conclusion, by addressing the key challenges of predictability and interpretability in choice modelling specification, my research advances the field of discrete choice modelling not only with transport applications but also in other domains such as economics and social sciences, offering innovative solutions that are methodologically robust
Correction: The support of early-career researchers in health professions education—an expert position statement (Frontiers in Medicine, (2025), 12, (1621194), 10.3389/fmed.2025.1621194)
No full text[This corrects the article DOI: 10.3389/fmed.2025.1621194.]
Smartwatches in childhood: a mixed blessing?—a scoping review
No full textBACKGROUND: Smartwatches are commonly used within the community, especially amongst its younger population. These devices have a wide range of capabilities, including measuring the heart rate, generating electrocardiographic traces, and issuing alerts when 'abnormal' activity is detected. This information has potential benefits but also potential risks if the health-related measurements lead to inappropriate clinical interventions. This study aimed to evaluate the current literature on the prevalence, perception and interpretation of smartwatches that document and record cardiac information as it impacts on children, adolescents and their parents. METHODS: We conducted a scoping review based on the principles of Arksey and O'Malley, which followed the scoping review checklist of the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR). RESULTS: The selection criteria yielded 29 papers. They reported that smartwatch usage in children ranged from 15% to 40%, depending on the country of residence. The number of children presenting with smartwatch-based heart concerns had increased, with many false positives and few true arrhythmia diagnoses. Although there was good accuracy of heart rate measurements, there were poor automated algorithms for heart rhythm classification for the paediatric population. In addition, a few studies reported paediatric smartwatch user anxiety arising from the information generated by the devices. CONCLUSIONS: The wearing of smartwatches has increased in children and adolescents. While they are able to record heart rates and provide corresponding electrocardiographic tracings virtually continuously and non-invasively, misinterpretation of the data arising from poor algorithms have led to increased healthcare presentations, as well as child and/or parental concern. There remains a need for ongoing education to understand the variability of the heart rate, especially in children. Furthermore, better algorithms for the interpretation of the information gleaned are required for this relatively well young population, so as to allay the anxiety that may be experienced. The issues related to medicolegal liability, privacy and cybersecurity remain to be resolved
Understanding the Role of Sports Injury Management by Australian Osteopaths: A Cross Sectional Survey of 992 Practitioners
No full textSport-related injuries are common presentations to primary care and hospital settings. Australian osteopaths practice mainly in private clinical settings in which the frequency of sport-related injury presentations, and how these injuries are managed, is unknown. The objective of the study was to describe the demographic, practice, and clinical management characteristics of Australian osteopaths who report often treating sport-related injuries. The study is a secondary analysis of data derived from the Australian osteopathy practice-based research network. Respondents indicated the frequency treating sports-related injuries in addition to other demographic, practice, and patient management characteristics. Backward logistic regression identified significant characteristics associated with often treating sport injuries. Over half (51%) of a nationally representative sample of Australian osteopaths reported treating sport-related injuries often. Those osteopaths who treat sports injuries often were likely to be male (p < 0.01) and utilise exercise prescription (OR2.34) and sports taping (OR5.99). Australian osteopaths who often treat sports-related injuries provide advice to patients and use exercise prescription more frequently than osteopaths who do not treat these injuries often. The data in the current work begin to explore how osteopaths manage sports-related injuries and highlights how they may be able to provide sports injury care for both recreational and elite sport populations
Integrating Neural and Computational Approaches to Recognition Memory: A Critical Examination of Theories and Assumptions
No full text© 2025 Jie SunThe ability to recognise objects from memory is fundamental to learning, decision-making, and daily functions. While cognitive models and neuroimaging methods have advanced our understanding of recognition memory, few attempted have directly linked neural signals to formal computational mechanisms. This thesis addresses this gap by integrating electroencephalography (EEG) with evidence accumulation models to investigate the neural basis of recognition memory decisions. Across three studies, we first re-evaluated key assumptions about established neural signatures and computational models of recognition memory, and then combined them in a joint modelling framework to uncover their functional relationships.
The first study re-evaluated widely held assumptions about the temporal dynamics of the Late Positive Component (LPC), a neural signal often linked to recollection. Using deconvolution techniques on a large EEG dataset (n = 132) with recognition memory tasks, we demonstrated that the LPC is more accurately characterised as time-locked to the memory decision response, rather than to stimulus onset. Additionally, we found that LPC amplitude differences previously attributed to recollection could be confounded by an overlapping confidence-related signal. These findings challenge the traditional interpretation of the LPC effects, suggesting a possible role of the LPC as a decisional variable, and offered analytical approaches to reduce the effects from confounding factors for future investigations.
The second study addresses an ad-hoc criticism of the across-trial drift rate variability assumption in the Diffusion Decision Model (DDM), which describes evidence strength differing across trials and modelled as a normal distribution. Specifically, with a neurally informed modelling framework, we examined whether this random distribution assumption could be explained by systematic information from experimental manipulations and EEG signals. While simulation results demonstrated that systematic trial-level information could in principle replace random drift rate variability, analysis of the empirical data revealed that such information only partially accounted for it. This suggests that drift rate variability estimates in recognition memory decisions may reflect not only fluctuations in evidence quality, but also additional latent processes not currently captured by the standard DDM.
In the third study, inspired by the mnemonic accumulator hypothesis of the parietal lobe, we investigated whether the LPC reflects a mnemonic evidence accumulation process, akin to the centro-parietal positivity (CPP) observed in perceptual decision-making. We used joint modelling to fit the DDM simultaneously to LPC amplitudes and behavioural data, applying specialised neural networks to perform likelihood-free inference. The results showed that LPC amplitudes, especially prior to the time of the response, were accounted for by trial-wise variation in drift rate estimates, with a stronger relationship for studied words compared to novel ones. These findings provided a fundamental reinterpretation of the LPC as a correlate of dynamic memory strength variable driving evidence accumulation in recognition memory decisions
Mitochondrial DNA variant detection in over 6,500 rare disease families by the systematic analysis of exome and genome sequencing data resolves undiagnosed cases
No full textVariants in the mitochondrial genome (mtDNA) cause a diverse collection of mitochondrial diseases and have extensive phenotypic overlap with Mendelian diseases encoded on the nuclear genome. The mtDNA is not always specifically evaluated in patients with suspected Mendelian disease, resulting in overlooked diagnostic variants. Here, we analyzed a cohort of 6,660 rare disease families (5,625 genetically undiagnosed [84%]) from the Genomics Research to Elucidate the Genetics of Rare diseases (GREGoR) Consortium, as well as other rare disease cohorts. Using dedicated pipelines to address the technical challenges posed by the mtDNA—circular genome, variant heteroplasmy, and nuclear misalignment—we called single nucleotide variants, small insertions/deletions, and large mtDNA deletions from exome and/or genome sequencing data, in addition to RNA sequencing data when available. Diagnostic mtDNA variants were identified in 10 previously genetically undiagnosed families (1 large deletion, 8 reported pathogenic variants, and 1 previously unreported likely pathogenic variant), as well as candidate diagnostic variants in a further 11 undiagnosed families. In one additional undiagnosed proband, detection of >900 heteroplasmic variants provided functional evidence of pathogenicity to a de novo variant in the nuclear gene POLG (DNA polymerase gamma), responsible for mtDNA replication and repair. Overall, mtDNA variant calling from data generated by exome and genome sequencing—primarily for nuclear variant analysis—resulted in a genetic diagnosis for 0.2% of undiagnosed families affected by a broad range of rare diseases, as well as the identification of additional promising candidates in 0.2%
Feasibility of a Progesterone-Modified Natural Protocol for Frozen Embryo Transfer: Protocol for a Pilot Cohort Study
No full textBACKGROUND: With the existence of various frozen embryo transfer (FET) methods currently used in the field of assisted reproductive technologies, the debate surrounding which of these is superior remains. All FET protocols aim to prime the endometrium and time embryo transfer during the window of implantation. Current methods include the true natural cycle FET (tNFET), modified natural cycle FET, artificial cycle FET, and ovulation induction. Each of these harbors, distinct advantages and disadvantages, namely, surrounding the timing of transfer and flexibility conferred through this process. More recently, a newer approach has been used whereby the need to monitor or trigger ovulation is circumvented, with luteal phase support commenced once a certain follicle diameter and endometrial thickness criteria are met but before ovulation. However, the research into this protocol has certain important limitations that our study seeks to address. OBJECTIVE: This study aims to assess the feasibility of a progesterone-modified natural cycle protocol for FET. The primary outcome will be the presence of a corpus luteum on ultrasound scans on the day of embryo transfer. The secondary outcomes will include the number of clinic visits required per patient undergoing the protocol, biochemical pregnancy rate, and clinical pregnancy rate. METHODS: We will conduct a prospective cohort study, recruiting 20 women undertaking FET at the Public Fertility Care of The Royal Women's Hospital in Melbourne, Australia. These women will be matched to a control group who have undergone the tNFET protocol within the preceding 12 months of the study start date. RESULTS: This project received ethics approval on July 17, 2024, with commencement of the study in September 2024, aiming for a duration of completion of 9 months. The completion of the follow-up and submission of the study for publication are anticipated for September 2025. CONCLUSIONS: After this preliminary study, the aim would be to progress to a noninferiority randomized controlled trial to compare the progesterone-modified natural cycle protocol for FET to the tNFET. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): PRR1-10.2196/66579
Management of Clostridioides difficile infection in patients with haematological malignancies and after cellular therapy: guidelines from 10th European Conference on Infections in Leukaemia (ECIL-10)
No full textClostridioides difficile infection (CDI) poses a significant challenge in patients with haematological malignancies (HM) and those undergoing cellular therapy such as haematopoietic cell transplantation (HCT) or CAR T-cell therapy. These patients have high rates of both colonization with Clostridioides difficile and diarrhoea due to non-infectious causes, leading to challenges with establishing diagnosis and optimal management of CDI, especially in the setting of molecular detection of toxin genes alone. Current severity criteria are of limited usefulness since underlying haematological disease and its treatment impact white blood count and inflammatory manifestations of severe CDI. Extensive exposure to antibiotics, profound microbiota damage and bidirectional relationship with gastro-intestinal graft-versus-host disease after transplant further complicate clinical management. Therefore, the 10th European Conference on Infections in Leukemia (ECIL-10) group comprehensively reviewed the literature (published 01/01/2010-15/09/2024) on the epidemiology, treatment and prevention of CDI, and formulated consensus recommendations for the management of CDI specific to this population. New definitions of proven, probable and possible CDI in this population were developed and proposed for use in clinical research to standardise reporting
Conducting descriptive epidemiology and causal inference studies using observational data: A 10-point primer for stroke researchers
No full textRoutinely-collected health data and emerging data-linkage capabilities provide researchers and clinicians with rich opportunities to answer important research questions by conducting observational studies. We provide stroke researchers with 10 important points to consider and implement to ensure the validity and interpretability of descriptive epidemiology and causal inference studies based on observational data. We discuss different types of observational studies and biases that may arise in such studies. We review types of causal effects and the use of Target Trial emulation and Directed Acyclic Graphs to improve validity of observational studies. We also illustrate appropriate and inappropriate use of covariate adjustment for the analyses of observational studies and review the methods for estimating the effects of treatments, interventions, and exposures in causal inference studies. Finally, we provide recommendations for clinical researchers and journal manuscript reviewers in stroke domain and beyond for the appropriate use and reporting of these methods