21684 research outputs found
Sort by
Applications of linearised oscillation models on the T2K experiment
Get PDFT2K is a long-baseline neutrino oscillation experiment measuring νμ disappearance and νe appearance with a baseline of 295 km and a narrow-band beam peaking in energy at around 0.6 GeV. This thesis presents a linearised oscillation probability model for the phase angle δCP, designed to complement standard analyses used in this experiment. We demonstrate that the framework provides a universal method for performing T2K sensitivity studies, and characterisation of systematics, that avoids almost all of the problems with commonly used metrics, most notably the dependence on the choice of trial point. Furthermore, we show how it restores the validity of well-known asymptotic approximations, allowing the likelihood to be described by a simple Gaussian form which could reduce or eliminate the need for computationally expensive tools such as MCMC or Frequentist toys sampling when quoting results at high significance
Simulation for design optimisation of Ge-on-Si single photon avalanche diodes
No full textAbstract not currently available
Advanced geotechnical laboratory investigation into the stiffness behaviour of a low- to medium-density chalk
No full textAbstract not currently available
Investigating the conditions that permit the survival of leukaemic cells in the central nervous system to construct an in vitro model and improve treatments for childhood leukaemia
No full textAbstract not currently available
Understanding human exposure to viral haemorrhagic fevers in Uganda: occupational, behavioural, and ecological factors
Get PDFIntroduction:
Uganda experiences frequent outbreaks of viral haemorrhagic fever viruses (VHFVs), placing healthcare workers (HCWs) and local communities at risk of exposure. Among these, Crimean Congo haemorrhagic fever virus (CCHFV) is a tick-borne zoonotic pathogen that can cause severe haemorrhagic disease with high fatality rates among hospitalised cases. Despite a rise in reported infections over the past decade, the true burden of CCHFV remains underestimated due to mild or misdiagnosed presentations. Understanding the complex interplay of occupational, behavioural, and ecological risk factors is essential for identifying high-risk populations and guiding effective interventions. Previous research has highlighted the importance of geographic variability in exposure risk, yet socioecological determinants remain poorly understood. This thesis aims to address these gaps and increase the knowledge around VHFVs with a focus on CCHFV.
Methods:
The body of work is based on three cohort studies designed to investigate exposure to CCHFV and other VHFVs in Uganda. Firstly, a case-control study was conducted among 639 HCWs and 714 age- and sex-matched community members, to understand occupational risk for VHFV exposure. The study sites comprised hospitals in Gulu, Arua and Kasese districts of Uganda. Serum was tested for Ebola virus (EBOV) and CCHFV seropositivity by ELISA and for Rift Valley fever virus (RVFV) by indirect immunofluorescence. Exposure risk factors were evaluated with a structured survey and analysed by multivariable logistic regression.
A qualitative investigation was next carried out to study human-animal-tick interactions through 24 focus group discussions (FGDs) and 31 key informant interviews (KIIs), in six environmentally and socioecologically diverse districts of Uganda. FGDs were conducted in groups of community leaders, men, women and teenagers. Medical doctors, veterinarians, traditional healers, district surveillance officers, and herdsmen were also interviewed as key informants. Data were translated into English, transcribed, and analysed using iterative categorisation.
The final quantitative cohort study used an analytical framework to estimate seroprevalence in the first four of six selected districts of Uganda as part of an interim analysis of the wider AVI study. 1,059 participants were recruited through multi-level randomisation and stratified by age. Serum samples were collected from each participant, and a structured survey was performed, which was informed by the preceding qualitative research. CCHFV antibody testing was carried out to estimate CCHFV exposure and force of infection (FOI) . Multivariable logistic regression was used to evaluate underlying risk factors.
Results:
Overall, seropositivity in the HCWs study was 16% for EBOV, 19% for CCHFV, and 2% for RVFV seropositivity. The highest odds of exposure were noted in Arua district for both EBOV (AOR = 9.01 [95% CI = 5.48-15.4]) and CCHFV (AOR = 4.67 [95% CI = 3.11-7.13]), around hospitals that had no previously documented cases of VHFVs. Overall, HCWs had a lower odds of EBOV exposure than community members (AOR = 0.37 [95% CI0.26-0.51]), as well as of CCHFV exposure (AOR = 0.42 [95% CI 0.31-0.57]). Homemakers and cleaners had the highest seropositivity for EBOV and CCHFV in the respective study groups.
Thirteen district clusters showed notable differences in climate, land use, proximity to wildlife, and subregional locations within Uganda. Six of these were selected for subsequent qualitative and quantitative cohort studies. Participants from both FGDs and KIIs described distinct living conditions and practices, highlighting regional variation.
The majority of the people that we interviewed as part of our qualitative study experienced tick bites, some as frequently as every day. Close contact with animals was common, including cohabitation, largely due to concerns about animal theft. Less frequent but notable practices included slaughtering animals for consumption or sacrifice, drinking blood, and interactions with wild animals during hunting. Slaughtering and butchering were reported if an animal was unwell or had died. Plucking and roasting engorged ticks for consumption was a practice described in the Kaabong and Arua districts of Northern Uganda.
The quantitative study highlighted varying estimated seroprevalence to CCHFV, ranging from 2.2% in Kaabong district to 18.2% in Kasese district. A multivariable analysis, including known risk factors for CCHFV transmission, revealed significant differences in CCHFV seropositivity between study locations (p = 0.002) and age groups (p < 0.001). The FOI showed an accumulation of seropositivity with age, suggesting constant exposure rather than isolated outbreaks.
Discussion
This PhD demonstrates that exposure to VHFVs in Uganda is extremely high, and is shaped by a complex interplay of ecological, occupational, and behavioural factors. In the HCWs study, seropositivity was highest for CCHFV (19%), followed by EBOV (16%) and RVFV (2%). The unexpectedly high odds of exposure in Arua district, where CCHFV has only very rarely been reported, strongly suggests the presence of mild and/or misdiagnosed cases. Elevated risk among homemakers and cleaners, within community members and HCWs respectively, points to occupational exposures that have been largely overlooked. Qualitative findings, including daily tick bites, animal cohabitation, and practices such as tick collection for consumption, underscore the need for context-specific evaluation of risk behaviours in Uganda’s diverse settings. These behaviours represent possible transmission routes for CCHFV and highlight the importance of future studies to quantify their contribution to infection risk, and to identify targeted and culturally appropriate interventions. The serosurvey revealed significant variation in estimated seroprevalence across surveyed districts (ranging between 2.2% and 18.2%), reinforcing previous findings that study location and, therefore, environmental and geographic factors are key drivers of exposure to CCHFV. These insights can support the identification of high-risk regions and guide targeted control strategies for CCHFV transmission, including the implementation of tick control measures and the prioritisation of future vaccine trials
Transformers and contrastive semi-supervised learning for medical image segmentation
Get PDFMedical Image Semantic Segmentation (MISS), the process of assigning a semantic label to each pixel in an image, is a foundational task in computational medicine, critical for quantitative diagnostics and treatment planning. However, developing robust MISS models faces two intertwined challenges. First, there is an architectural dilemma: Convolutional Neural Networks (CNNs), like U-Net, excel at learning local features but are limited by their receptive fields, failing to capture global context essential for segmenting organs with large deformations. Conversely, Vision Transformers (ViTs) effectively model long-range dependencies but lack the inductive biases of CNNs, leading to poor generalization on the small datasets typical in medicine without extensive pre-training. Second, the prohibitive cost and expertise required for creating pixel-level annotations create a severe data scarcity bottleneck. While Semi-Supervised Learning (SSL) aims to mitigate this by leveraging unlabeled data, existing methods often fail to learn high-level semantic relations and are susceptible to confirmation bias from noisy pseudo-labels, class imbalance, and suboptimal contrastive sample selection.
This thesis presents a comprehensive investigation to systematically address these challenges, delivering a cohesive suite of novel deep learning frameworks. The contributions are four-fold:
First, to resolve the architectural trade-off, this work introduces CS-Unet, a pure Transformer network built upon a U-Net-like architecture. Its core innovation is the Convolutional Swin Transformer (CST) block, which integrates convolutions directly within the Multi-Head Self-Attention and Feed-Forward Network modules. This design imbues the Transformer with inherent localized spatial context and strong inductive biases, enabling it to efficiently learn both local and global features. Without pre-training, CS-Unet outperforms existing Transformer and CNN-based models on multi-organ and cardiac datasets, achieving state-of-the-art performance with fewer parameters.
Second, to address data scarcity, a novel Multi-Scale Cross Supervised Contrastive Learning (MCSC) framework for SSL is developed. MCSC jointly trains CNN and Transformer models, using a cross-teaching paradigm where each network provides pseudo-labels for the other. Crucially, it moves beyond simple output consistency by applying a contrastive loss to feature maps at multiple scales, enforcing hierarchical semantic consistency. To handle the class imbalance endemic to medical imaging, a class-prevalence-aware loss is used to ensure features for infrequent classes are learned robustly.
Third, to fortify SSL against noisy pseudo-labels, a certainty-guided contrastive learning strategy is proposed. This approach mitigates the impact of inaccurate pseudo-labels by using a certainty metric to guide the selection of samples for contrastive learning. The framework’s computational efficiency is enhanced through novel sampling strategies that select a few representative samples for contrasting, and a negative memory bank is used to increase sample diversity and eliminate dependence on batch size.
Fourth, this thesis introduces a new paradigm for SSL by leveraging external anatomical priors through the Contrastive Cross-Teaching with Registration (CCT-R) framework. CCT-R is the first method to integrate spatial registration transforms into the learning process. It features two novel modules: a Registration Supervision Loss (RSL), which uses transforms between labeled and unlabeled volumes to generate an additional, highly reliable source of pseudo-labels, and Registration-Enhanced Positive Sampling (REPS), which uses registration to identify anatomically-corresponding positive pairs across volumes for contrastive learning.
Overall, these contributions provide a powerful toolkit that significantly alleviates the annotation bottleneck in medical AI. The proposed methods demonstrate state-of-the-art performance on challenging segmentation benchmarks, delivering a pathway to develop accurate, data-efficient models for real-world clinical applications and opening new avenues for research into fusing geometric priors with semantic segmentation
Novel probes for multimodal imaging at the nanoscale
Get PDFTip enhanced Raman spectroscopy is a technique which allows the user to retrieve topographic, morphological and chemical information of a sample with single molecule resolution, far beyond the Raman spatial resolution of conventional techniques. Its usefulness extends into many scientific fields, such as medicine, electronics and material science. Conventional probes used to conduct TERS , where metal is deposited onto commercial AFM tips, tend to be made of silver. This is due to silver having strong plasmon resonances in the visible and near-infrared regime, alongside low optical losses and there are well established processes for making them. The issue with these probes is that they are commonly deposited one at a time and silver is known to grow a sulphide layer. This sulphide film inhibits the electric field, shifting the plasmon resonance and limits the lifetime of these probes to 24-48hrs. The use of a conventional AFM tip as a substrate for the plasmonic film is also problematic since the coupling of the far field radiation to the tip plasmon occurs close to the tip and is facilitated by local roughness, leading to critical and irreproducible alignment requirements for the illuminating beam. The issue of tip contamination and lifetime may be eliminated by the use of gold instead of silver, at the expense of plasmonic performance. The use of a long-lived metal allows the use of more elaborate coupling methods based on plasmonic gratings or slot couplers fabricated by focused ion beam milling or self-aligned plasma processing. Both techniques of silver growth and gold coupler tip fabrication are irreproducible however, they do not have nm resolution placement or techniques to ensure the same tip is made over again using plasma or chemical etches. The work-around for this is to use a focused ion beam with the benefit of very high resolution (nm), but each probe has to be milled one at a time. A lot of effort goes into making these tips, where if one fails due to over/under etch or other reasons, another metal deposition or plasma processing has to occur as the tip is scrapped.
A method for the wafer scale fabrication of novel TERS-AFM probes is presented. The probes are based on the use of a grating to couple light from free space into the dielectric of the tip at some distance from its apex, the grating allowing for the use of fixed – angle illumination at a known position far from the tip. Waveguide to surface plasmon coupling occurs near the tip and plasmons are then concentrated to the tip by a triangular metal structure. The fabrication was successfully completed, although operation at useful wavelengths was precluded by poor reproducibility of the probe dielectric thickness. Development work on the fabrication of scanning electro-chemical microscopy probes, where insulator adhesion to the metal electrode is a key issue, has also been been conducted.
To test the TERS probes, novel calibration samples were made. These samples employed statistical and correlation alignment strategies with an e-beam lithography tool to produce 70nm thick gold dots and dimers with separation distances of 1nm and above (1nm increments) in the x and y direction. Once coated with Raman Active molecules this sample will be useful for testing of TERS probes, allowing the study of separation distance vs. enhancement for different probe archetypes. A topography free sample with interdigitated electrode active areas has also been developed for the test of Scanning Electrochemical Microscope probes. This sample removes the introduction of topographic artefacts from a SECM scan, whilst retrieving useful information on the electro-chemical nature of the probes.
After extensive and rigorous testing, probes failed to exhibit local plasmonic enhancement at the available wavelength of the TERS systems. The lack of enhancement was confirmed by extensive processing of the acquired data using baseline correction and adaptive smoothness penalised least squares (asPLS) methods. For comparison of the SERS and TERS spectra the data was also intensity normalised and cosmic rays removed. This is attributed to poor control of dielectric deposition thickness preventing phase matching of the guided light to the surface plasmon polariton: Possible solutions are presented and their practicality discussed
Diversity and dynamics of the Atlantic salmon gill microbiome in marine production and its relationship to AGD
No full textAbstract not currently available
Data sources:
Access Repository Dashboard
Do you manage Open Research Online? Become a CORE Member to access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard! 👇