136986 research outputs found
Sort by
REMOTE SENSING IMAGES AND COMPUTER VISION TECHNIQUES FOR URBAN BUILDING FOOTPRINT EXTRACTION
No full text© 2025 Bipul NeupaneAccurate and reliable building footprint datasets are fundamental to the effective monitoring, planning, and management of urban environments. With the increasing availability of high-resolution satellite and aerial imagery, semantic segmentation using deep learning has become the standard method for generating such datasets. However, despite significant progress, several critical challenges persist that limit the accuracy, scalability, and generalisability of current extraction methods. This thesis investigates four knowledge gaps that impact urban building footprint extraction as identified from a systematic and comprehensive literature review. They are label displacement errors caused by off-nadir imagery, lack of cross-domain robustness coming from the differences in different geographic and imaging conditions, incomplete training data, and irregularly segmented outputs of building footprints.
To address the four knowledge gaps, the thesis presents several key novel contributions. First, a systematic survey of 165 research articles is conducted, dissecting methods, datasets, training practices, and evaluation approaches within urban feature extraction, and providing new insights into deep learning and remote sensing-based challenges. An extensive benchmarking of convolutional neural network (CNN)-based image segmentation models, encoder–decoder networks, and training hyperparameters is undertaken to identify optimal configurations for building footprint extraction. To mitigate label displacement errors, a fine-tuning-based transfer learning (FTL) approach is proposed, enabling robust adaptation of segmentation models to noisy and misaligned datasets. Additionally, a novel multilabel learning framework is developed, using Vision Transformer (ViT)-based architectures to simultaneously learn oblique-view features such as footprints, roofs, and shapes, improving footprint extraction performance in off-nadir images. To improve cross-domain robustness, a model-agnostic phase-wise gradual fine-tuning (PGFiT) method is introduced for optimising transferability under concept, covariate, and prior shifts across diverse urban datasets. To train accurate models despite incomplete training datasets, this thesis proposes Open Mutual Learning (OML). It is a collaborative framework that integrates incomplete open-source and commercial building footprint datasets. Furthermore, an end-to-end building boundary regularisation method is developed to replace conventional post-processing for producing regularised building shapes.
Together, these innovations address the identified knowledge gaps and advance the literature of urban building footprint extraction by offering robust, scalable, and transferable solutions. The methods developed in this thesis not only improve extraction accuracy across challenging conditions but also contribute to the broader field of remote sensing and computer vision, supporting future applications in urban analytics, disaster assessment, and large-scale geospatial monitoring
Generation and selection of training events for surrogate flood inundation models
No full textThe destructive and life-threatening nature of flood events calls for fast and accurate methods to predict dynamic flood behaviour. Data-driven surrogate models have been developed to quickly predict flood inundation, though their accuracy relies on the available flood information for model training and validation. Flood observations are rarely available at high spatial and temporal scales, and thus computationally expensive high-resolution hydrodynamic (high-fidelity) models are often used to generate training data through simulation of selected flood events. Given finite resources, only a limited number of events can be simulated using a high-fidelity model. However, there is no established approach for selecting representative and informative flood events to ensure that the surrogate model is robustly trained. In this study, a novel systematic approach for selecting flood events for the training of surrogate flood inundation models is introduced. The approach generates a large set of candidate events using a computationally efficient low-resolution hydrodynamic (low-fidelity) model and then selects training events based on the simulated spatial-temporal inundation depths of the candidate events. The approach is used to train surrogate models to predict flood inundation in three distinct case studies with different boundary conditions and topographies. The results show robust performance of the surrogate models developed with RMSE<0.23 m when applied to new unseen events, which is similar to the accuracy achieved when using all available candidate events for training. This means the proposed training event selection approach reduces the computational costs of generating training data by up to 97% as fewer high-fidelity model simulations are needed, highlighting the computational advantage of the approach. Although this study focuses on surrogate models for the prediction of flood inundation dynamics, the new approach could easily be used for the development of surrogate models in other fields
Randomized Phase Ib Clinical Trial of DB-020 Intratympanic Injections to Reduce High-Dose Cisplatin Ototoxicity
No full textPURPOSE: This study evaluated DB-020, a formulation of thiosulfate for intratympanic (IT) injection, in patients receiving high-dose cisplatin chemotherapy. METHODS: This randomized phase Ib clinical trial enrolled patients older than 18 years from five centers in Australia and the United States scheduled for at least three cycles of cisplatin and total cumulative exposure of ≥280 mg/m2. Patients received IT DB-020 (at a dose level of either 12% or 25%) in one ear and placebo in the other, once every 3 or 4 weeks, within 3 hours before receiving cisplatin. The primary end points were safety and tolerability, and the secondary end points included ototoxicity measured by air conduction audiometry. Ototoxicity was defined by American Speech-Language-Hearing Association criteria. RESULTS: Twenty-two patients with a median age of 55.1 years were randomly assigned and received a mean total cumulative cisplatin dose of 255 mg/m2. Mean number of cisplatin cycles was 2.3. Twenty patients had both baseline and follow-up audiometry. Ear pain of short duration was common after IT injection. There were no persistent tympanic perforations and no serious adverse events in the category of ear and labyrinth disorders. A progressive reduction in patient numbers was observed at each cycle due to patients ceasing cisplatin treatment. DB-020 treatment did not affect plasma thiosulfate concentrations. Ototoxicity after cisplatin administration was significantly more common in placebo-treated ears than in DB-020-treated ears (DB-020 v placebo, P = .0027). The incidence of ototoxicity (250-8,000 Hz) was 85.0% in placebo-treated ears, and 54.5% and 22.2% in ears treated with DB-020 12% and DB-020 25%, respectively. CONCLUSION: DB-020 IT injections were tolerated by patients and showed meaningful reductions in cisplatin ototoxicity
Impact of a Momentary Mindfulness Intervention on Rumination, Negative Affect, and their Dynamics in Daily Life
No full textUNLABELLED: Rumination and negative affect are mutually reinforcing experiences. Their dynamic relation can confer vulnerability to psychopathology. Cultivating mindfulness has been proposed to buffer against such downward spirals of negativity. However, it remains unclear whether practicing mindfulness in daily life causally impacts rumination, negative affect, and their dynamics. We investigated this using a micro-randomized intensive longitudinal trial. Participants (N = 91) were prompted eight times per day for 10 days using a smartphone app. At each prompt, participants were randomized to complete a brief mindfulness intervention or an active-control task and then reported levels of rumination and negative affect. Results of dynamic structural equation models showed that the mindfulness intervention led to lower levels of rumination and negative affect but that it had no reliable impact on their dynamics. Thus, cultivating mindfulness in daily life may be a promising approach for decreasing rumination and negative affect but not their dynamical relation. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s42761-024-00291-9
ISA3: a 3-dimensional expansion of instance space analysis
No full textAbstract
The experimental validation of algorithms depends strongly on the characteristics of the test set used. Ideally, such a set should exhibit diverse characteristics that challenge an algorithm and are present in real-world problems. An approach to examine the diversity and representativeness of a test set is Instance Space Analysis, which uses a 2D projection to visualise the test set while identifying the strengths and weaknesses of competing algorithms. However, this has the limitation of discarding potentially useful information while crowding out the space as more features and algorithms are considered. This paper describes an extension of Instance Space Analysis into 3D, which retains a higher degree of information while maintaining the explainability of the visualisation by finding the rotations that maximise the linear trends. In addition to the expansion to 3D, a new algorithm for identifying portfolio footprints is introduced, offering a more robust and reliable method for identifying footprints in both 2D and 3D instance spaces. As a case study, we present a performance analysis of unsupervised anomaly detection methods, subject to changes in the normalisation technique. The results demonstrate the advantages by identifying regions of strength of algorithms previously thought to be underpowered
Advancing equity: the urgent need to include trans and gender diverse people in cervical cancer prevention in Australia
No full textPatient and Surgical Factors Associated With Long-Term Mortality Outcomes Up to Fifteen Years After Total Hip and Knee Arthroplasty: An Australian Orthopaedic Association National Joint Replacement Registry Study
No full textBACKGROUND: Mortality rates following total hip arthroplasty (THA) and total knee arthroplasty (TKA) arthroplasty display distinct temporal patterns often attributed to patient selection bias, perioperative optimization, and comorbidities. Understanding these mortality patterns is essential for epidemiological and health economic longitudinal modeling. METHODS: We conducted a national registry-based cohort study in Australia using data from 1999 to 2022, examining primary and revision THA and TKA procedures for osteoarthritis. We assessed patient factors (age, sex, body mass index, and American Society of Anaesthesiologists score), and surgical factors (procedure, fixation, bearing surface, and implant volume) in relation to long-term mortality. Standardized mortality ratios were calculated by comparing observed and expected deaths based on national mortality rates. RESULTS: Our study included 540,181 THA and 880,036 TKA procedures. Temporal trends in mortality rates were observed, with a reduction in mortality rate observed up to seven years for both primary THA and primary TKA after the index procedure and an increased mortality rate observed thereafter. All patient factors were associated with differences in mortality rates, with younger (age range, 45 to 49 years) patients for primary TKA demonstrating the strongest association with mortality excess (15 years; standardized mortality ratios 2.02; 95% confidence interval 1.66 to 2.46). Revision procedures were associated with higher mortality rates compared to their respective primary procedures at all time points. CONCLUSIONS: Our study finds noncausal associations between patient and surgical factors and mortality up to fifteen years following THA and TKA for osteoarthritis in Australia. These findings are crucial for calibrating epidemiological and economic models and enhancing the precision of longitudinal outcome predictions for arthroplasty patients. While limitations exist, our study informs clinical practice, healthcare policies, and future research in arthroplasty surgery on a national scale, with potential relevance to similar populations worldwide. LEVEL OF EVIDENCE: Level III, therapeutic study
Spatial and temporal dynamics of the leaf area index (LAI) of selected tree species in urban environments
No full textKnowledge about a tree's leaf biomass, leaf area or leaf area index (LAI) are important prerequisites for deriving precise estimates of its ecosystem services. However, data availability on LAI is limited especially for urban trees, and influences including site conditions, tree size and vitality or seasonal changes are hardly known and usually not considered. The aims of this study are (i) to present a comprehensive database of LAI data over tree size for 15 urban tree species, and (ii) to evaluate seasonal LAI data for over 100 trees of six species in relation to species traits and location with hemispherical photography. Further, (iii) data on leaf emergence were analyzed and compared with data from the German Weather Service (DWD) to derive broader ecological trends. Our results showed that LAI increases with size. The highest LAI was observed at maximum stem diameter, with Tilia cordata showing the highest (maximum LAI 4.7) and Gleditsia triacanthos the lowest values (maximum LAI 2.4). Furthermore, we found significant differences in seasonal LAI development and leaf emergence, influenced by species traits like light requirements, sprouting type, wood anatomy and foliage density. Leaf patterns followed the typical course of a steep increase from branch area index (BAI) values of 0.3 in spring to a maximum LAI of 4.0 in summer and autumn/winter values of 0.5 BAI/LAI. The influence of soil sealing on LAI was less pronounced, albeit statistically significant. The study highlights that the hemispheric image method can be applied easily to individual urban trees and can support a precise calculation of ecosystem services. However, the data processing has some weaknesses, and the results should be adjusted with a correction factor
Effects of an eLearning course for patients on osteoarthritis knowledge and pain self-efficacy in people with hip and/or knee osteoarthritis: A randomised controlled trial
No full textObjective: Evaluate a knee/hip osteoarthritis eLearning course for patients. Methods: Randomised controlled trial. 124 community volunteers with knee/hip osteoarthritis were randomised to either i) a 4-week self-directed eLearning or ii) an electronic osteoarthritis pamphlet (control). Primary outcomes: change in knowledge (Osteoarthritis Knowledge Scale (OAKS)) and pain self-efficacy (Arthritis Self-Efficacy Scale (ASES pain subscale)) over 5 weeks. Secondary outcomes: fear of movement, exercise self-efficacy, osteoarthritis illness perceptions, physical activity levels, and use of physical activity/exercise, weight loss, pain medication, and health professional care seeking to manage joint symptoms. Results: 117(94 %) participants (mean (SD) age, 67.1(8.8) years; 91(77.8 %) female) provided 5-week primary outcomes. At 5-weeks, eLearning group showed greater improvements in osteoarthritis knowledge (mean difference 5.3(95 % CI 2.5,8.2), < 0.001), which was sustained at 13-weeks (4.6(2.1,7.0), < 0.001). There were no between-group differences in pain self-efficacy. Between-group differences for exercise self-efficacy and osteoarthritis illness perceptions at 5-weeks, and fear of movement and use of weight loss to manage joint symptoms at 13-weeks, favoured eLearning group. Conclusions: eLearning produced immediate and sustained improvements in osteoarthritis knowledge but not pain self-efficacy compared to a typical osteoarthritis education intervention (information pamphlet). Practice implications: Self-directed interactive eLearning is an effective method to educate patients about hip/knee osteoarthritis and its management
Regulation of airway fumarate by host and pathogen promotes Staphylococcus aureus pneumonia
No full textStaphylococcus aureus is a leading cause of healthcare-associated pneumonia, contributing significantly to morbidity and mortality worldwide. As a ubiquitous colonizer of the upper respiratory tract, S. aureus must undergo substantial metabolic adaptation to achieve persistent infection in the distinctive microenvironment of the lung. We observed that fumC, which encodes the enzyme that converts fumarate to malate, is highly conserved with low mutation rates in S. aureus isolates from chronic lung infections. Fumarate, a pro-inflammatory metabolite produced by macrophages during infection, is regulated by the host fumarate hydratase (FH) to limit inflammation. Here, we demonstrate that fumarate, which accumulates in the chronically infected lung, is detrimental to S. aureus, blocking primary metabolic pathways such as glycolysis and oxidative phosphorylation (OXPHOS). This creates a metabolic bottleneck that drives staphylococcal FH (FumC) activity for airway adaptation. FumC not only degrades fumarate but also directs its utilization into critical pathways including the tricarboxylic acid (TCA) cycle, gluconeogenesis and hexosamine synthesis to maintain metabolic fitness and form a protective biofilm. Itaconate, another abundant immunometabolite in the infected airway enhances FumC activity, in synergy with fumarate. In a mouse model of pneumonia, a ΔfumC mutant displays significant attenuation compared to its parent and complemented strains, particularly in fumarate- and itaconate-replete conditions. Our findings underscore the pivotal role of immunometabolites in promoting S. aureus pulmonary adaptation.10.1038/s41467-025-62453-