Sydney eScholarship
Not a member yet
31878 research outputs found
Sort by
Mechanosensing and routes of glioblastoma dissemination
No full textGliomas are the most common malignant primary brain tumours, with high-grade gliomas (HGG), such as glioblastoma multiforme (GBM), exhibiting strong resistance to standard treatments like surgical resection and chemoradiation. Despite decades of research, patient outcomes remain poor, underscoring the need to better understand therapeutic failure.
The tumour microenvironment (TME) plays a crucial role in driving tumour heterogeneity, invasion, and therapy resistance. While biochemical factors of the TME have been widely studied, recent focus has turned to its mechanical properties and how these influence GBM migration and adaptation. The failure of promising preclinical therapies in clinical settings is increasingly attributed to gaps in understanding the TME’s biomechanics, mechanotransduction pathways, and interactions with surrounding cells.
This thesis investigates the impact of cell-cell interactions on HGG migration using a 3D spheroid model with patient-derived GBM cells cultured on biomechanically relevant substrates. Findings revealed distinct and shared migration patterns across cell lines, shaped by cell-cell and cell-substrate dynamics, potentially mediated by focal adhesions (FA), and largely independent of GBM subtype. In vivo studies using a mouse xenograft model with the same cell lines further highlighted the roles of EGFR and FAK signalling in guiding GBM invasion. A systematic review of the brain’s viscoelastic properties under health, disease, and treatment conditions revealed a lack of longitudinal data on mechanical changes post-therapy—an essential gap for advancing precise treatments.
Collectively, this work emphasises the importance of TME mechanics in GBM progression and therapy resistance, offering insights for the development of more effective interventions for this devastating disease
Few-Shot Relational Learning on Knowledge Graphs: Towards Model Adaptation and Generalization
No full textKnowledge graphs are essential for structuring vast information and enabling advanced inference in
applications like QA, web search, and recommendation. However, their inherent incompleteness
limits reasoning, driving research in relational learning to infer missing facts via expressive
representations.
Traditional embedding-based methods succeed but rely on large-scale data. Many relations,
however, have few triplets, making generalization difficult. Few-shot relational learning addresses this
by learning from minimal examples.
Despite progress, challenges remain: (1) overemphasis on entity embeddings over relational
structures, (2) assumptions violating permutation invariance, (3) task-isolated learning that misses
transferable patterns, and (4) reliance on relational data while ignoring semantic knowledge.
This thesis systematically tackles these challenges. First, we introduce a hierarchical framework that
jointly models entity, triplet, and context information, ensuring permutation invariance and improved
generalization. Second, we propose a meta-learning framework with a Mixture-of-Experts model for
relational prototypes, balancing global generalization and local adaptability. Third, we integrate
semantic knowledge via a prompted meta-learning framework, enhancing inference of unseen
relations and paving the way for large language models.
Additionally, we release pretrained semantic embeddings for benchmark datasets to support future
research. This thesis lays a foundation for more robust and semantically enriched knowledge graph
relational learning, advancing intelligent knowledge graphs
Catalysts and systems for carbon dioxide electro-conversion
No full textEnergy consumption is the primary cause of human-induced global warming, which contributes to 75.6% of global emissions. In this context, carbon dioxide (CO2) capture and conversion become the favored option via chemical, thermal, biological, electrochemical, and photochemical methods. Within these options, electrochemical carbon dioxide reduction reaction (CO2RR) powered by renewable electricity provides a sustainable avenue to convert CO2 into valuable fuels and achieve negative carbon emissions. Additionally, the conversion of intermittent renewable electricity into the chemicals is beneficial for storage and transport, avoiding energy waste and offering diverse energy usage scenarios. In this thesis, I focus on catalyst engineering and system design to achieve efficient CO2 electrolysis. I synthesized a coordination polymer catalyst Cu(OH)BTA with homogenized, single-site Cu active sites, which is found to be stable and efficient for CO2RR to C2+ products (Chapter 2). Then I extended the categories of coordination polymer catalysts with tunable Cu electronic states through ligand modification, revealing a volcano-shaped correlation between the binding strength of *CO intermediate and the C–C coupling efficiency (Chapter 3). To further overcome the carbonate issue in above alkaline CO2 electrolysis, I investigated the feasibility of alkali-metal-cation-free electrolytes for CO2RR, which employed non-metal polymeric cations as electrolytes and clarified the relationship between CO2RR performance and interfacial water structures (Chapter 4). This thesis explores catalyst engineering and system design for efficient CO2 electrolysis, focusing on the design of coordination polymer catalysts and the use of alkali-metal-cation-free electrolytes to enhance the selectivity and stability of CO2RR
High temperature effects on wheat yield, grain quality, and pollen lipids
No full textWheat (Triticum aestivum L.) contributes ~20% of global dietary calories and 21% of daily protein intake, underpinning global food security. However, high-temperature (HT) stress increasingly threatens wheat yield and quality by accelerating development, impairing reproduction, and altering grain composition. This thesis investigates HT effects on wheat grain traits and reproductive resilience using both field and controlled-environment experiments across genotypes differing in heat tolerance.
Field trials were conducted in 2019 and 2020 at Narrabri (NSW), Horsham (VIC), and Merredin (WA), employing early and late sowing to vary HT exposure during grain filling. Cultivars included Berkut and Flanker (heat-tolerant), Sokoll and Suntop (moderately tolerant), and Cobra (sensitive). Analyses encompassed yield, thousand kernel weight (TKW), protein fractions, starch, fibre, RVA profiles, mineral content, and phytate levels. Statistical analyses included ANOVA and Tukey’s HSD.
HT generally reduced yield and TKW, while crude protein increased, especially in sensitive cultivars, though glutenin-to-gliadin ratios declined. Starch content and pasting properties deteriorated under HT, with some cultivars (e.g., Cobra, Sokoll) exhibiting significant reductions in peak and breakdown viscosities. Crude fibre increased in some cases, likely due to accelerated lignification. HT also influenced mineral uptake and phytate concentration, shaped by cultivar and soil type, with Berkut retaining higher mineral levels.
A controlled-environment experiment assessed pollen lipid composition in four cultivars under three temperature regimes (22/15°C, 35/22°C, and 40/22°C). HT reduced both saturated and unsaturated fatty acids, particularly in Cobra, correlating with poor pollen viability. In contrast, Flanker and Suntop maintained lipid homeostasis, indicating superior thermotolerance
Anterior mitral leaflet interventions and valvular-ventricular mechanoenergetics
No full textPatients with functional mitral regurgitation and poor left ventricular function face an increased risk of morbidity and mortality. Although mitral repair techniques exist, valve replacement is often favored due to reduced reoperation risk. Preserving the subvalvular apparatus during replacement is crucial for maintaining left ventricular function, though most methods retain only the posterior leaflet. Modifying or removing the anterior mitral leaflet, typically to avoid left ventricular outflow tract obstruction, may impair left ventricular function—an area not well understood.
This study developed a normothermic beating-heart ovine model to examine the acute effects of mitral interventions on left ventricular function across three experiments:
1. Experiment 1 evaluated the effects of different bioprosthetic interstrut distances on the native anterior mitral leaflet movement in five sheep. Wider inter-strut distance subtending the anterior leaflet partially protected against left ventricular outflow tract obstruction.
2. Experiment 2 involved 14 sheep with mechanical mitral valve insertion while retaining the native mitral valve. Releasable snares reefed and released the anterior leaflet to the annulus, showing that reefing altered left ventricular hemodynamics, reduced contractility, and increased left ventricular sphericity.
3. Experiment 3 in six sheep evaluated splitting the anterior leaflet and shortening the annulo-papillary distance. These interventions acutely impaired left ventricular contractility and hemodynamics.
Overall, the research demonstrated that commonly used anterior leaflet interventions can significantly affect left ventricular function. The findings underscore the importance of preserving the entire valvular-ventricular apparatus and provide a foundation for improving valve designs and surgical techniques
Universal Health Care Delivery Mitigates Socioeconomic-Related Risk for Adverse Outcomes in Hospitalised Patients : Lessons from the COVID-19 Pandemic in Australia
No full textObjectives: Internationally, socioeconomic disadvantage is related to severe outcomes of COVID-19. We investigated the impact of socioeconomic disadvantage on infection rates, hospitalisation, and in-hospital outcomes for COVID-19 with standardised medical care.
Design: Retrospective cross-sectional study.
Setting: SARS-CoV-2 PCR-confirmed patients, ≥18 years old, admitted to a major public hospital between January 2020 and December 2021.
Main outcome measurements: Severe COVID-19 outcomes were defined by a composite outcome of in-hospital death or other critical complications. A generalised linear regression model of demographic features, co-existing conditions, and socioeconomic status [Socioeconomic Index for Area (SEIFA)] was used to determine the risks of the composite outcome.
Results: Of 797,343 individuals aged ≥18 in the health district, 50,906 (6.4%) were PCR-positive, and 1,962 were hospitalised. Compared with the whole health district population, infected individuals were younger (median [interquartile range] age 35 [25-48] years vs 42 [31-58] years) and from areas with the greatest socioeconomic disadvantage (34.4% vs 20%; both p<0.0001). Hospitalised patients were older, with more females compared to the PCR-positive group (46 years [33-61], 53.5%, respectively; p<0.001), and 51.2% were from postcodes with greatest socioeconomic disadvantage (p<0.0001). The composite outcome occurred in 11.5%, with an in-hospital mortality of 3.8%. Higher risk of the composite outcome was observed in males (OR 1.72, 95% CI [1.26-2.42], p <0.001), patients aged ≥ 65 years (OR 6.96, [3.3-14.6], p <0.001), those with ≥4comorbidities (OR 2.67, [1.54-4.63], p <0.001), and unvaccinated patients (OR 1.57, [1.05-2.38], p < 0.05). The risk of composite outcome did not increase with socioeconomic disadvantage (OR 0.97, [0.68, 1.42], p = 0.64).
Conclusion: In the absence of capacity restraints, socioeconomic disadvantage was not associated with severe in-hospital outcomes in a well-resourced care environment despite the increased rates of infection and hospitalisation. This highlights the impact of universally accessible, standardised, protocolised, high-quality in-hospital care in reducing the risk of adverse in-hospital outcomes in socioeconomically disadvantaged patients
Application of Deep Learning in Image Processing
No full textDeep learning has revolutionized computer vision, achieving remarkable success in complex visual tasks such as image classification, segmentation, and object detection. This thesis explores advanced deep learning techniques for image processing, with a particular focus on remote sensing imagery. Specifically, this research addresses two critical challenges: (1) achieving high segmentation accuracy in scenarios with limited labeled data and (2) integrating multi-modality data into model training.
To tackle these challenges, this work proposes innovative solutions leveraging semi-supervised learning and multi-modality learning. By employing consistency learning and advanced data augmentation techniques, the proposed approaches effectively utilize unlabeled data, significantly boosting segmentation accuracy.
Furthermore, integrating complementary data modalities, such as spectral and spatial information, enhances model robustness and overall performance. Experimental results on benchmark datasets validate the effectiveness of these methods, demonstrating their potential for real-world applications, including environmental monitoring, urban planning, and disaster management.
The primary contributions of this thesis include advancing the theoretical understanding of semi-supervised and multi-modality learning in remote sensing segmentation, developing novel methodologies to address data scarcity, and providing practical frameworks that are applicable across various domains. However, limitations related to scalability and generalizability highlight avenues for future research, such as exploring dynamic augmentation strategies, advanced fusion mechanisms, and extensions to other fields like medical imaging.
This research provides a comprehensive framework for overcoming segmentation challenges in remote sensing, delivering significant advancements in deep learning-based image analysis
A Deep Learning Framework for Real-Time Cancer Targeting in Radiation Therapy
No full textGrowing evidence highlights the detrimental effects of underdosing tumours and overdosing organs at risk during high dose radiation therapy, emphasising the need for precision in treatment delivery. Tumour motion, due to normal physiological processes, is one factor that compromises accuracy. To address this, real-time motion management technologies have been developed to continuously monitor the tumour position. These technologies, though effective, are either prohibitively expensive or require the implantation of fiducial markers. These barriers were highlighted in a recent international survey which revealed that 71% of responding centres wish to implement real-time motion management for another treatment site but are limited by resources and capacity. This thesis presents the first large-scale proof-of-principle for x-ray-based markerless segmentation in globally available radiation therapy systems. It provides an important step towards making real-time motion management treatments accessible for all patients, eliminating the need for expensive, dedicated equipment or fiducial marker implantation. The proposed markerless approach relies solely on x-ray images acquired during treatment which in principle, covers the majority of linear accelerators, thus overcoming the resource and capacity barriers to real-time motion management. The first two studies investigate a deep learning framework for markerless segmentation of the prostate and pancreas head, demonstrating high accuracy across a large patient cohort. The third study explores using residual contrast agents from liver cancer chemotherapy as a surrogate for real-time motion monitoring, showing successful tracking in patients with liver cancer. A software application was also developed for ground truth data labelling. The thesis concludes with key findings and future directions for deep learning-based markerless tumour tracking
Error Analysis of Bosonic Codes
No full textQuantum computers have been shown to be able to perform computations that are thus far thought to be infeasible
through classical computation. This is achieved by taking advantage of certain properties and phenomena unique to
quantum systems. Unfortunately, engineering particular quantum systems that are useful for computation is difficult
and there is still a long road towards building a robust, large-scale quantum computer.
A key approach to building quantum computers that are robust to noise is the development of Quantum Error
Correcting codes. In recent years, there have been a surge of progress in the development and realisation of bosonic
codes, which are quantum error correcting codes that encode quantum information into the infinite-dimensional
Hilbert space of an oscillator. Due to its large Hilbert space, bosonic codes are able to protect quantum information
using a single, or a few physical systems.
This thesis will take two different but complementary approaches to the development of bosonic codes and bosonic
code qubits. The first will be a numerical study of errors affecting bosonic codes within a fault-tolerant system.
Specifically, we introduce the concatenated Bacon-Shor rotation-symmetric bosonic code and evaluate the
entanglement fidelity of a teleportation based error correction scheme against pure loss errors.
In the second approach, we study the recently proposed gyrator qubit by Rymarz et al., which realises a bosonic code
known as the Gottesman-Kitaev-Preskill (GKP) code in its low energy eigenspace. We support this study with a
Wentzel-Kramers-Brillouin (WKB) analysis of the Zak transformed Hamiltonian of the qubit. This allows us to derive
the low-energy spectrum and eigenstates of the qubit and derive matrix elements corresponding to transitions between qubit states when the system is weakly coupled to a thermal environment