136986 research outputs found
Sort by
Therapeutic approach to difficult-to-treat multidrug-resistant enterococcal infections
No full textDifficult-to-treat (DTR) enterococcal infections, particularly those caused by multidrug-resistant Enterococcus faecium and Enterococcus faecalis, pose significant clinical challenges due to limited treatment options and high rates of treatment failure, compounded by a paucity of new antimicrobial agents in the development pipeline. Despite advances in understanding resistance mechanisms and in vitro synergistic antibiotic combinations, robust clinical data to guide therapy for severe or DTR enterococcal infections remain limited. This review synthesizes available evidence to inform optimal management strategies, including drug selection and dosing, while highlighting areas needing further research. Given the ongoing threat posed by multidrug-resistant enterococci, we emphasize the importance of gathering robust clinical data to guide best practices for managing these difficult-to-treat infections
Arsenic in Soil: A Critical and Scoping Review of Exposure Pathways and Health Impacts
No full textArsenic (As) in soil, such as mining waste, is a concern for communities with legacy contamination. While the chronic health effects of As exposure through drinking water are well documented, the association between As in soil and population-wide health impacts is complex, involving factors like soil accessibility, soil properties, and exposure modes. This review summarizes evidence of associations between As in soil and human health, as well as biomarker and bioaccessibility evidence of exposure pathways. Fourteen studies were included in the final analysis. Reviewed studies reported associations between As in soil and birth outcomes, neurological effects, DNA damage, and cancer. Some of these health outcomes are not known to be linked to As in drinking water and were reported over a range of soil concentrations, indicating inconsistencies. Higher soil As concentrations are associated with higher As in human biospecimens, suggesting direct and indirect soil ingestion as primary exposure pathways. The subpopulations more likely to be exposed include younger children and those involved in soil-based activities. Future research should focus on standardized epidemiological studies, longitudinal studies, soil exposure and mitigating factors, combined exposure biomarker studies, the behavior of the different As species, soil dose related to bioavailability/bioaccessibility, and effects with other elements
Population-based longitudinal study over two decades of Candida and Candida-like species bloodstream infection reveals gender and species differences in mortality, recurrence and resistance
No full textBACKGROUND: The global burden of bloodstream infection (BSI) due to Candida, and species previously classed as Candida (Candida-like species) is substantial. Recent emergence of Candida auris, fluconazole-resistant Candida parapsilosis and echinocandin-resistant Nakaseomyces glabratus emphasise the importance of global and regional surveillance. METHODS: Blood cultures with growth of Candida/Candida-like species in Queensland, Australia (population ≈ 5 million) over a 20-year period (1 January 2000-31 December 2019) were retrospectively identified. Clinical, microbiological and outcome information was obtained from state-wide databases. Cox proportional and Fine-Gray subdistribution hazard models were used to construct hazard ratios for 30-day all-cause case fatality and 1-year recurrence, respectively. RESULTS: A total of 2586 episodes (2420 patients) of Candida/Candida-like bloodstream infection (Ca-BSI) were identified; 249 episodes (9.5%) were in children. Candida albicans and C. parapsilosis complex reduced in frequency, whilst N. glabratus and Candida dubliniensis increased during the study. Of 1836 isolates tested, fluconazole (3.2%) and echinocandin (0.7%) resistance rates were low, with a decrease in fluconazole resistance observed from the first half of the study period to the latter half (4.5% versus 2.2%, P<0.01). Overall, 30-day all-cause mortality (21%) was unchanged: C. parapsilosis complex (aHR 0.44, 95% CI 0.32-0.60) was associated with decreased mortality, while C. tropicalis (aHR 1.35, 95% CI 0.95-1.93) was associated with an increase. Only 3.1% episodes demonstrated recurrence of Ca-BSI within one year. Presence of uncommon Candida species (aSHR 6.60, 95% CI 2.99-14.56) and an endovascular source of infection (aSHR 4.42, 95% CI 1.87-10.46) were associated with recurrence, while male gender (aSHR 0.57, 95% CI 0.35-0.92) was protective. Resistance to fluconazole (3.2% vs 3.5%, P=0.58) and echinocandins (0.6% vs 2.0%, P=0.05) was higher in recurrent Ca-BSI episodes. Females had a higher rate of fluconazole resistance (4.1% versus 2.4%, P=0.02). CONCLUSIONS: Our study highlights important shifts in causative species and resistance patterns of Ca-BSI which impacts clinical management. Antifungal resistance rates were low overall. The identification of new modifiable and non-modifiable risk factors for recurrence and mortality provides opportunities to examine new strategies to improve patient outcomes
Breathing Cycle-Aware Segmentation for Patient-Ventilator Asynchrony Detection
No full textPatient–ventilator asynchrony (PVA) is a significant challenge in mechanical ventilation, affecting approximately 25% of intensive care unit patients and increasing the risk of lung and diaphragm injury. Segmenting breathing cycles from long ventilation waveforms is essential for the reliable detection of PVA events. However, existing segmentation methods present several limitations: manual annotation is time-consuming; fixed-length window and rule-based segmentation methods lack adaptability to varying respiratory patterns; and supervised deep learning (DL) segmentation methods require large amounts of labelled data for training. To address these issues, we propose an unsupervised breathing cycle-aware segmentation method tailored for PVA detection. Leveraging the quasi-periodic nature of ventilation waveforms, the proposed segmentation method integrates frequency-adaptive clustering, periodicity hints validation, and dynamic segmentation to identify breathing cycle boundaries. We evaluate the proposed breathing cycle-aware segmentation method on a real-world dataset from Austin Health, Melbourne, Australia, where it outperforms baseline approaches on five out of six evaluation metrics. Furthermore, classification experiments using two state-of-the-art DL-based classification models confirm that accurate segmentation of breathing cycles enhances PVA detection performance. In the future, the proposed breathing cycle-aware segmentation method could be integrated into ventilation systems to support clinical decision-making and improve patient care
Prevalence and incidence of falls in older adults with cancer: a systematic review with meta-analysis
No full textPURPOSE: To determine the prevalence and incidence of falls and fall-related injuries in community-dwelling older adults with a diagnosis of cancer and examine whether falls prevalence varies with specific cancer characteristics. METHODS: A systematic search of five databases was conducted. Studies that included community-dwelling adults with a mean age ≥ 60 years with a current or past diagnosis of cancer and that reported data on the prevalence and/or incidence of falls and/or fall-related injuries were included. Prevalence and incidence rates of falls were pooled for meta-analysis, and meta-regression was used to investigate associations between cancer characteristics (e.g. cancer type and cancer treatment received) and prevalence of falls. RESULTS: Fifty-seven studies with sample sizes ranging from 51-146,959 participants were included. The pooled prevalence of older adults with cancer who fell in the last 6 months (25%; 95% CI 19%, 32%) and 12 months (29%; 95% CI 24%, 34%) was similar. Subgroup analysis showed that the pooled prevalence of falls for older adults with breast cancer was higher (26%; 95% CI 22%, 30%) compared to those with prostate (14%; 95% CI 9%, 20%) or colorectal cancer (13%; 95% CI 11%, 16%). CONCLUSIONS: The overall prevalence of falls amongst community-dwelling older adults with cancer is relatively similar compared to the general older adult population, noting that fall events may have been under-reported. IMPLICATIONS FOR CANCER SURVIVORS: Falls are common amongst older adults with cancer, but the link between cancer characteristics and exposure to falls risk requires further investigation to better understand the risk factors specific to cancer survivors
Correction: Family and parenting factors are associated with emotion regulation neural function in early adolescent girls with elevated internalizing symptoms (European Child & Adolescent Psychiatry, (2024), 33, 12, (4381-4391), 10.1007/s00787-024-02481-z)
No full textTracing Singularities in Deep Learning: From Toy Models to Sequence Models via Singular Learning Theory
No full text© 2025 Zhongtian ChenDespite the empirical success of deep learning models such as transformers, their theoretical foundations remain insufficiently understood. This thesis investigates how Singular Learning Theory (SLT), a mathematical framework designed to analyze models with non-identifiable parameters and degenerate loss landscapes, can be used to understand the learning dynamics of neural networks.
This thesis comprises two main parts. In the first part, it applies SLT to the Toy Model
of Superposition (TMS), a simplified neural network designed for the research in mechanistic interpretability. We show that the principle of internal model selection proposed in SLT indeed happens in TMS. Moreover, the stochastic gradient descent (SGD) exhibits dynamical transitions in line with SLT’s predictions, governed by a tradeoff between loss and model complexity measured by local learning coefficients (LLCs). These results highlight the practical utility of the LLC estimator, which reveals phase transitions in
training.
The second part addresses the limitations of the LLC estimator in large-scale settings. In
more complex models, where SGD trajectories are smoother and transitions are less distinct, the interpretation of LLC estimates becomes ambiguous. To clarify its theoretical
foundation, we develop a Hilbert space formalism for sequence models. By using tensor
decompositions, we show that, under certain assumption, the LLC estimates reflects the
geometry of a coarse-grained approximation of the true data distribution. This provides
a theoretical foundation for understanding the LLC estimates and offers a principled
justification for its usage in large-scale settings such as transformers
Sources, transients, and the Simons Observatory
No full text© 2025 Justin Patrick ClancyThe Simons Observatory (SO) marks the beginning of a new chapter in studies of the cosmic microwave background (CMB).
With its improved sensitivity and versatility, SO broadens the millimetre-wave sky to a vast array of astrophysical goals beyond traditional cosmology.
This thesis explores diverse scientific applications of the SO large aperture telescope, with a focus on astrophysical sources and transients -- from the polarization properties of Galactic cold clumps, to the detection of rapid flaring astrophysical transients, and systematic biases in galaxy cluster measurements from relativistic corrections to the thermal Sunyaev-Zel`dovich (SZ) effect.
We begin by presenting the first measurement of the mean-squared polarization fraction of Galactic cold clumps using Planck satellite data, providing statistical insight into the magnetic field structures of star-forming regions in our Galaxy.
With an 11 sigma detection of polarization ([4.79+-0.44]x10^-4), we forecast the capabilities of SO in mapping this polarized emission from cold clumps, finding that SO should detect ~430 cold clumps directly in polarization at >5 sigma, a 200-fold increase over Planck results.
Next, we present a novel approach to blindly detect rapid astrophysical transients in time-ordered data from the SO large aperture telescope.
We estimate the detection efficiency of our matched-filter approach as a function of peak flux density for a range of flare-like lightcurve timescales.
For events flaring in just 0.5 seconds, we achieve >90 % detection efficiency for peak fluxes as low as 800 mJy at 90 GHz and 1150 mJy at 150 GHz in simulated large aperture telescope observations.
We also demonstrate robust source localization and accurate flux recovery, with multiple blind detection capabilities of a single event across detector wafers.
Our pipeline, now being integrated into SO's broader time-domain framework, will enable the large aperture telescope to systematically monitor the rapidly varying millimetre sky.
Finally, we assess the impact of relativistic corrections to the thermal SZ effect in the context of galaxy cluster detection.
Using simulated Compton-y maps painted from dark matter halo catalogues, we quantify an average suppression of -4.32+-1.48 % of the integrated Compton-Y in high signal-to-noise clusters, reaching -6.69+-2.77 % for the hottest systems (T_Y>10 keV).
These corrections prove significant relative to SO noise levels and should be accounted for to avoid biasing cosmological constraints with the large cluster samples expected from SO
Diaspora Humanitarianism: Centring Transnational Relational Care Practices in Crisis Response
No full textThis paper investigates diasporas’ mobilization in response to humanitarian crises, in comparison to the ‘traditional’ humanitarian sector. We develop a conceptual framework through which we analyze eight Australia‐based diasporas’ responses to humanitarian crises. We find that diaspora humanitarian interventions are premised on transnational reciprocal relationships and care practices that bind people together. Diaspora crisis responses significantly diverge from traditional humanitarianism, which often focuses on the provision of goods without scope for reciprocal engagement. We argue that the universalist ethical principles that guide the traditional international humanitarian system (i.e., humanity, impartiality, neutrality and independence) are not well‐suited to understand diaspora responses to crises. We propose that the ontologies of feminist ethics of care and of relational humanitarianism are best suited to recognize diaspora humanitarianism, as they allow for an integrated analysis of diverse diaspora transnational relational care practices, the spheres in which these occur and their underlying motivations
Pore-scale Analysis of Fluid and Particle Transport Behaviour in Granular Materials
No full text© 2025 Jie QiEmerging challenges in geotechnical engineering, such as dam safety, geothermal energy exploitation, and soil erosion, demand a deeper understanding of the micro-scale behaviour of granular materials. These materials, characterised by their complex pore structures, govern critical hydraulic and mechanical properties in multi-physics processes. Despite advancements, significant gaps remain in understanding the relationships between microstructural features and macroscopic responses, particularly under coupled hydraulic and mechanical conditions. Limitations in imaging resolution, numerical accuracy, and predictive models hinder progress in accurately characterizing fluid flow, hydraulic anisotropy, and internal erosion in granular materials.
This thesis addresses these challenges by developing a multidisciplinary framework integrating advanced numerical simulations, experimental techniques, and data-driven methods. Specifically, the study combines the Lattice Boltzmann Method (LBM) and Discrete Element Method (DEM) to model fluid-solid interactions, incorporating irregular particle shapes derived from CT scans. The work advances pore network analysis by leveraging complex network theory to characterise microstructural features such as pore connectivity and hydraulic conductance. Experimental validation is conducted through 3D printing of granular assemblies and permeability tests using a custom-designed permeameter. The CFD-DEM simulations of internal erosion in gap-graded soils further extend this framework, employing convolutional autoencoders to detect erosion initiation and reveal the micro-scale mechanisms leading to macroscopic failure.
The results demonstrate significant insights into fluid flow, hydraulic anisotropy, and internal erosion mechanisms. LBM-DEM simulations highlight the critical role of particle shape and microstructural parameters in governing hydraulic conductivity and mechanical stability. Pore network analysis uncovers the impact of connectivity and heterogeneity on fluid transport. Experimental validation confirms numerical predictions, and the coupled CFD-DEM simulations successfully model erosion processes while the autoencoder provides a robust tool for anomaly detection.
By linking microstructural features to macroscopic response, this thesis bridges critical research gaps, offering novel methodologies to characterise fluid flow in soils, and to predict and mitigate risks in geotechnical engineering applications. The findings advance the scientific understanding of granular materials, contributing to the development of safer and more sustainable engineering solutions