143174 research outputs found
Sort by
Decoding auditory EEG responses to speech via deep neural networks
No full textNeural speech tracking – the phenomenon by which a listener’s neural activity synchronises with a speech stimulus – is represented in the listener’s electro-encephalogram (EEG). In the continuous speech paradigm, participants listen to extended, naturalistic narratives whilst their EEG is recorded; various aspects of their perception and cognition can subsequently be decoded from these measurements. In this thesis, new methods based on deep learning are employed for the purpose of auditory EEG decoding. Accurate auditory EEG decoders could find clinical applications in the objective diagnosis of hearing disorders, or in cognitively-steered hearing aids which monitor the focus of the listener’s auditory attention and adapt their behaviour accordingly.
We first employed non-linear models based on deep neural networks for the purpose of reconstructing the speech envelope from the listeners’ EEG measurements. These decoders performed significantly better than their linear counterparts, and could also function as auditory attention decoders. However, they did not generalise well between different listeners, a property which would be valuable in translational applications. We then turned to a different decoding problem, the match-mismatch problem, which was part of the ICASSP 2023/2024 Auditory EEG Decoding Signal Processing Grand Challenge series. Inspired by known responses to speech – envelope tracking and speech-related frequency-following responses – we developed highly accurate, challenge-winning decoders, which this time generalised well between participants and even across datasets. Finally, we applied some of these techniques in an auditory attention identification setting using EEG signals recorded from a wearable ear-EEG device. Whilst auditory attention could be identified with significance using this device, the results demonstrate that real-world auditory attention decoding remains a challenging goal. Taken together, our results highlight the strengths and weaknesses of using deep learning for auditory EEG decoding, and highlight important topics for future research.Open Acces
Computer vision for carbonate core classification
No full textCore data is a fundamentally important resource in characterising the geological subsurface. Globally, cores provide a record of diverse past geologic, climatic, and biological processes that have operated over multiple timescales. Leveraging insights from core images presents a pivotal challenge due to the largely subjective nature of traditional core description methods, inherent complexities and variability in geological data, and the paucity of significantly large, labelled datasets. Deep learning presents a data-driven approach to deriving predictive models from observational data.
With the evolution of machine learning, and growing awareness of the importance of big data, new automated workflows using deep learning algorithms present alternative approaches to the extensive time and labour requirements of traditional, manual core descriptions. Focusing solely on carbonate rocks, this thesis aims to establish foundations for applying deep learning to core interpretation. First, convolutional neural networks (CNNs) are used to classify carbonate textures from core images, optimising a workflow for lithological prediction across datasets of different magnitudes. To address the limiting factor of insufficient data, Generative Adversarial Networks (GANs) are applied to generate synthetic core images, thus filling sampling gaps and reducing overfitting. The performance of synthetic data is then compared with traditional data augmentation methods. Lastly, CNN-based object detection algorithms are applied to identify and quantify multiple types of carbonate grain, with results benchmarked against human interpretations to evaluate the efficacy of the method.
This approach uses computer vision models, such as CNNs and GANs, to enable streamlined geological interpretation and data-driven insights. As geoscience data availability increases through new datasets and the digitalisation of historical records, these methods can be further developed. The frameworks proposed in this thesis offer a pathway for advancing deep learning-based carbonate classification and have the potential for broader application to cores from diverse formations and lithologies.Open Acces
Modulation of droplet trajectory using oscillating electric fields
No full textThe use of sinusoidal electric forces to modulate the location of charged fuel droplets in a channel flow configuration is studied as a potential technology to achieve control over fuel–air mixing. Trajectories of a single non-evaporating droplet are first investigated using both an analytical solution based on a linearised drag model and results obtained with a non-linear model for the drag force. Non-dimensional quantities are also introduced to generalise the problem and provide useful information for technology development. Then, the effects of oscillating electric fields on the dispersion and stratification of fuel vapour are investigated using large-eddy simulations. Kerosene is used as fuel, which is relevant for aviation applications. Results show that oscillating electric fields perpendicular to the bulk flow can significantly enhance the spreading of both droplets and fuel vapour, an effect that is more evident with decreasing angular frequency of the oscillation and decreasing flow bulk velocity. The effects of oscillating electric forces on the spreading of droplets become negligible above a given threshold of the angular frequency of the oscillations. Non-dimensional analysis shows that this threshold depends on both the intensity of the electric force and the gas-phase properties, in particular the dynamic viscosity. The distribution of droplets in the channel is characterised by regions of preferential concentration, where fuel vapour mass fraction tends to be higher. Results also demonstrate that the time-averaged local fuel–air mixing is significantly affected by the period of oscillations, evaporation timescale and residence time in the channel. The present investigation further supports the proposal of using electric forces to improve fuel–air mixing in liquid-fuelled combustors, opening up new possibilities for fuel-preparation systems based on electrohydrodynamics
Iron triphosphine complexes for the fixation of dinitrogen
No full textAmmonia, which is essential for global fertiliser production and a potential zero-carbon fuel alternative, is one of the world's most produced chemicals. The Haber-Bosch process, discovered in the early 20 th century, initiated industrial-scale ammonia production from dinitrogen and dihydrogen. However, despite its century-long existence, optimisation has been stagnant, and the process still requires harsh conditions, consuming 2% of global energy and contributing 1.8% of worldwide carbon dioxide emissions annually. Homogeneous inorganic chemistry research aims to understand dinitrogen fixation at a fundamental level, potentially leading to optimised ammonia production under milder conditions. This thesis explores the chemistry of novel Fe, Co, and Ni triphosphine complexes to elucidate the effects of the ligand structure and metals on nitrogen activation. Chapter 1 introduces industrial and biological nitrogen fixation methods, the fundamental principles of dinitrogen-metal binding, and an overview of nitrogen functionalization using iron catalysts. Chapter 2 explores the synthesis and metalation of multiple triphosphine ligands on Fe, Co, and Ni using a new methodology to produce diallyl phosphine and phosphine chloride precursors. Chapter 3 investigates the formation of isostructural dinitrogen complexes using Fe, Co, and Ni, furnished with the same triphosphine ligand. Chapter 4 discusses catalytic dinitrogen silylation and compares the reactivities of the Fe, Co, and Ni complexes. Chapter 5 examines the reactivity of an iron triphosphine complex with other small molecules including carbon dioxide, dihydrogen, and carbon monoxide.Open Acces
Hardware acceleration of Bayesian graph neural networks
No full textBayesian graph convolutional networks (Bayes-GCNs) have shown potential to quantify calibrated uncertainty along with their prediction for various graph-based learning applications. However, the inference of Monte Carlo (MC) Dropout-based Bayes-GCNs requires repeated MC sampling to perform uncertainty prediction, putting a heavy burden on hardware performance. To address this performance challenge, this work proposes an algorithm and hardware co-optimization approach to accelerate Bayes-GCNs. The first contribution of this thesis is to exploit three categories of sparsity to optimize the performance of Bayes-GCNs, including precision sparsity, layer sparsity, and sample sparsity. To automate this optimization, we propose a systematic framework to explore structured sparsity for given Bayes-GCNs while maintaining the same level of accuracy. The second contribution of this thesis is a reconfigurable FPGA-based hardware accelerator tailored for Bayes-GCNs. Several hardware optimization techniques are introduced to improve performance, including bit fusion to optimize the operation of single-bit data, compressed adjacency matrix storage to manage sparsity effectively, an optimized resource allocation strategy to improve resource utilization, and an optimized Bernoulli sampling strategy to improve data reuse. The third contribution of this thesis is the comprehensive evaluation of our approach based on widely-used datasets such as Cora, PubMed, and Flickr. The results demonstrate that our design can achieve up to 398 times faster performance than CPU implementations and up to 4.2 times faster than GPU implementations. Additionally, our design shows improvements in energy efficiency and latency compared to existing GCN accelerators.Open Acces
Vaginal microbiome dysbiosis and sexually transmitted infections correlate with concentrations of immunoglobulin isotypes in human cervicovaginal mucus: insights into HIV-1 transmission
No full textIntroduction: Little is known about the relationship between antibody isotype in cervicovaginal mucus (CVM) and the local microenvironment and how this impacts HIV-1 transmission at the female genital mucosa.
Methods: In a cohort of 139 adult women in Kenya, we measured antibody isotypes in CVM and describe their associations with local pH, serum concentrations of estrogen and progesterone, and sexually transmitted infections (STIs), including HIV-1.
Results: We found that immunoglobulin G2 (IgG2) was the most abundant and IgG4 was the least abundant in the CVM. Overall, IgG1 concentrations were significantly lower in CVM samples from women with bacterial vaginosis (BV) compared to those without BV. Among women with BV, IgG1 concentrations declined further as vaginal pH increased, suggesting possible pH-mediated degradation. We also report negative associations of BV status with IgG3 and IgG4. In addition, infection with Mycoplasma genitalium and Neisseria gonorrhoeae was positively associated with concentrations of IgA and IgM, respectively. We also found the relationship between antibody isotype and subclasses with HIV-1 viral mobility in vitro. IgG3 concentrations negatively correlated with CAP045 HIV-1 mobility and IgG1 concentrations negatively correlated with the mobility of the 92TH023 recombinant HIV-1 strain upon VRC01 depletion. These observations point towards a potentially protective role for IgG1 and IgG3 in trapping certain HIV-1 strains in the CVM.
Discussion: Importantly, our study builds on previous work, providing a potential mechanism by which BV and STIs may modulate immunoglobulin isotype and subclass content in the CVM. These results highlight the need for proper treatment of BV and other STIs, as this could impact the effectiveness of HIV-1 vaccines targeted at enhancing specific immunoglobulin responses in the cervicovaginal mucosa
Characterising spatiotemporal dynamics of arboviruses: mapping risk and assessing the impact of meteorological and environmental conditions on transmission
No full textArboviruses cause drastic health and economic burdens and are a leading cause of death and disability globally. There is an urgent need to better understand arbovirus transmission drivers to inform surveillance and control of disease. This thesis explores the associations between arbovirus risk and meteorology and environment, and assesses how well these drivers can reconstruct the transmission dynamics of different arboviruses across a range of settings. Furthermore, local hotspots of arbovirus risk are identified and annual patterns in transmission intensity are characterised to explore substantial spatiotemporal heterogeneities in arbovirus transmission patterns. The results of this thesis highlight the health impact of extreme meteorological conditions on local arbovirus transmission, which has important implications for epidemiological modelling and climate planning.Open Acces
Solar harvesting maximisation in spectral-splitting photovoltaic-thermal systems via integrated radiative cooling
No full textThe mismatch between the energy distribution of the solar spectrum and the spectral response of photovoltaic (PV) cells, along with the elevated operating temperature of the cells, especially in optically concentrating systems, can result in a significant loss of electrical performance and acceleration of ageing. Reducing PV cell temperature and improving spectrum utilisation are two key approaches to improve PV efficiency. As a zero-emission passive cooling technology, radiative cooling (RC) developed recently has promising potential for solar cells. How to integrate RC to make full use of the solar spectrum under highly concentrated conditions, and the quantitative relationship between concentration area and RC area, are still unclear so far. In this study, a spectral-splitting multi-cell photovoltaic-thermal system is integrated with RC technology to maximise solar spectrum utilisation, specifically exploring the relations between concentration and RC to achieve maximum utilisation efficiency. There exists an optimal concentrating ratio that maximises PV efficiency for a given RC area, and the optimal concentrating ratio increases as the RC area and solar cell's bandgap energy increases. In a multi-cell system, the coordinated distribution of RC area among cells can improve the overall efficiency relative to the uniform distribution of RC area. An increasing number of cells can effectively improve efficiency while reducing the required RC area allocated to each cell, providing a more feasible approach for RC application under highly concentrated conditions. The efficiency of this system with a low concentration ratio can achieve the theoretical maximum efficiency of a conventional PV system with a high concentration ratio. The efficiency limit of the proposed solar system can reach 73%, demonstrating the significant potential of the proposed concept in practical applications
UK medical cannabis registry: an updated clinical outcomes analysis of patients with post-traumatic stress disorder
No full textBackground
Cannabis-based medicinal products (CBMPs) are a potential treatment for post-traumatic stress disorder (PTSD), but their long-term efficacy and safety need further investigation. This study assessed the changes in health-related quality of life (HRQoL) and adverse events in PTSD patients prescribed CBMPs.
Research design and methods
This observational cohort study included PTSD patients enrolled on the UK Medical Cannabis Registry for 18 months or longer. Changes in PTSD-specific symptoms (IES-R), anxiety (GAD-7), sleep quality (SQS), and general HRQoL (EQ-5D-5 L) were assessed at 1, 3, 6, 12, and 18 months.
Results
In 269 patients, significant improvements in PTSD symptoms, anxiety, sleep quality, and HRQoL were observed at all follow-up points (p < 0.001). On multivariate logistic regression, male gender (OR = 0.51; 95% CI:0.28–0.94; p = 0.034) was associated with a reduced chance of reporting improvements in IES-R. Adverse events were reported by 70 (26.02%) patients, with insomnia (n = 42, 15.61%) and fatigue (n = 40, 14.87%) being the most common.
Conclusions
CBMPs were associated with improvements in PTSD symptoms, anxiety, sleep, and HRQoL at up to 18 months. Although the study’s observational nature limits causal conclusions, these findings support further assessment of medical cannabis.
Trial registration
This is an observational study and is not registered as a clinical trial
