University of Surrey - Research Portal

Background

Chronic lymphocytic leukaemia (CLL) has a heterogeneous lifelong course. While some patients never require treatment, most experience intermittent periods of active monitoring with other time points in active treatment. Most patients experience significant symptoms which negatively impact their quality of life (QoL). Although physical activity and exercise may help manage symptoms, it is unclear what disease-related factors drive the physical inactivity observed in people with CLL.

Methods

This study explored physical activity among people with CLL and assessed differences and relationships in treatment stage, symptoms, quality of life, and preferences for physical activity using an online questionnaire.

Results

A total of 128 individuals with CLL [66 M/62F: mean age 67 ± 9.1 years (range 38–91 years)] completed the questionnaire. Those who are being/have been treated (N = 55) exhibited worse QoL (p = 0.018) and lower engagement in higher levels of physical activity (p = 0.045) when compared to their treatment naïve (N = 73) counterparts. Both groups had similar symptomology, with fatigue (∼77%) and insomnia (∼55%) being the most reported and associated with less likelihood of being physically active. Physically active participants reported better QoL (p = 0.020), physical functioning (p = 0.003) and role functioning (p = 0.020) as well as lower levels of fatigue (p = 0.036), pain (p = 0.017) and symptom burden (p = 0.026) compared to those who were insufficiently active. Although 79% of respondents wanted to engage in exercise programs for their CLL, 70% reported never receiving exercise guidance from their healthcare professionals.

Conclusion

These findings highlight a significant need for targeted interventions to increase physical activity, likely improving QoL, in people with CLL. Furthermore, there is considerable interest from the CLL community in receiving exercise guidance; however, factors such as treatment status and symptomology should be considered when developing CLL-specific exercise programs.

Accurate estimation of soil properties is crucial for optimizing agricultural practices and promoting sustainable resource management. Hyperspectral imaging provides a non-invasive means of quantifying key soil parameters, but effectively utilizing the high-dimensional hyperspectral data presents significant challenges. In this paper, we introduce HyperSoilNet, a hybrid deep learning framework for estimating soil properties from hyperspectral imagery. HyperSoilNet leverages a pretrained hyperspectral-native CNN backbone and integrates it with a carefully optimized machine learning (ML) ensemble to combine the strengths of deep representation learning with traditional ML techniques. We evaluate our framework on the Hyperview challenge dataset, focusing on four critical soil properties: potassium oxide, phosphorus pentoxide, magnesium, and soil pH. Comprehensive experiments demonstrate that HyperSoilNet surpasses state-of-the-art models, achieving a score of 0.762 on the challenge leaderboard. Through detailed ablation studies and spectral analysis, we provide insights on the components of the framework, and their contribution to performance, showcasing its potential for advancing precision agriculture and sustainable soil management practices.

Background/Objectives: Both traumatic and stressful events, including major life changes, may contribute to post-traumatic stress symptoms (PTS), often associated with anxiety and depression. Feelings of loneliness may influence these relationships, whilst social support seems to mitigate the effects of stressful events on mental health. Our study thus aimed to evaluate the mediating role of loneliness in the relationships between PTS and both anxiety and depressive symptoms among university students. Methods: The data were from the CAMPUS study (0058642/21; FHMS 20-21 157), a survey on university students’ mental health in Italy and the UK. Using a logit model, mediation analyses were carried out to test whether the relationships between PTS and both anxiety and depressive symptoms might be mediated by loneliness. A path analysis was then performed to jointly test the associations between the Impact of Event Scale—Revised (IES-R)’s subscales and clinical domains. Results: Positive associations were found between PTS and both anxiety (p < 0.001) and depressive symptoms (p < 0.001). However, loneliness mediated approximately 22% of the effect of the PTS on anxiety symptoms (indirect effect: 1.04, 95% CI: 0.59; 1.48, p < 0.001) and approximately 33% of the effect of the PTS on depressive symptoms (indirect effect: 1.81, 95% CI: 1.22; 2.39, p < 0.001). Furthermore, the path analysis indicated associations between the IES-R’s hyperarousal subscale and both anxiety (coeff.: 0.34, p < 0.001) and depressive symptoms (coeff.: 0.27, p < 0.001). Conclusions: Along with the associations between PTS and both anxiety and depressive symptoms, our findings highlight the key role of loneliness in both these associations. Targeted interventions to reduce loneliness, especially for students exposed to traumatic events, may ultimately improve their mental health.

Nonbinary narratives are not just defined by genderqueer narrators or focalisers; they also move beyond binaries in the way they shape form and language. The study of nonbinary narratives therefore necessitates nonbinary thought and methods. In this essay, I therefore identify some of the ways in which texts can be read and analysed in a nonbinary manner. Building on queer, feminist and trans narrative studies, and on Kanaka Maoli and Two-Spirit understandings of what constitutes binary-defying narratives, I present an overview of current nonbinary narrative research. In order to illustrate what a nonbinary approach to narration might look like, I present three brief case studies of the use of singular ‘they’ in Rae Spoon’s novel Green Glass Ghosts (2021), the nonbinary form of Sara Taylor's The Lauras (2017) and genre transgressions in Joshua Whitehead's Making Love with the Land (2022).

Realistic, high-fidelity 3D facial animations are crucial for expressive avatar systems in human-computer interaction and accessibility. Although prior methods show promising quality, their reliance on the mesh domain limits their ability to fully leverage the rapid visual innovations seen in 2D computer vision and graphics. We propose VisualSpeaker, a novel method that bridges this gap using photorealistic differentiable rendering, supervised by visual speech recognition, for improved 3D facial animation. Our contribution is a perceptual lip-reading loss, derived by passing photorealistic 3D Gaussian Splatting avatar renders through a pre-trained Visual Automatic Speech Recognition model during training. Evaluation on the MEAD dataset demonstrates that VisualSpeaker improves both the standard Lip Vertex Error metric by 56.1% and the perceptual quality of the generated animations, while retaining the controllability of mesh-driven animation. This perceptual focus naturally supports accurate mouthings, essential cues that disambiguate similar manual signs in sign language avatars.

Gloss-free Sign Language Translation (SLT) has advanced rapidly, achieving strong performances without relying on gloss annotations. However, these gains have often come with increased model complexity and high computational demands, raising concerns about scalability, especially as large-scale sign language datasets become more common. We propose a segment-aware visual tokenization framework that leverages sign segmentation to convert continuous video into discrete, sign-informed visual tokens. This reduces input sequence length by up to 50% compared to prior methods, resulting in up to 2.67× lower memory usage and better scalability on larger datasets. To bridge the visual and linguistic modalities, we introduce a token-to-token contrastive alignment objective, along with a dual-level supervision that aligns both language embeddings and intermediate hidden states. This improves fine-grained cross-modal alignment without relying on gloss-level supervision. Our approach notably exceeds the performance of state-of-the-art methods on the PHOENIX14T benchmark, while significantly reducing sequence length. Further experiments also demonstrate our improved performance over prior work under comparable sequence-lengths, validating the potential of our tokenization and alignment strategies.

Tracking micro-unmanned aerial vehicles (micro-UAVs) in low signal-to-noise ratio (SNR) environments poses significant challenges due to their weak radar cross-section (RCS) and the inherent limitations of traditional Detect-Before-Track (DBT) radar algorithms. This paper proposes a novel Track-Before-Detect (TBD) approach based on a Markov Chain Monte Carlo-Enhanced Particle Filter (MCMC-EPF), leveraging 4D Frequency Modulated Continuous Wave (FMCW) MIMO radar. By directly processing unthresholded multi-frame 4D-FFT radar data, the method achieves joint detection and tracking, effectively preserving weak target information that is typically lost in DBT methods. Experimental results on real radar data demonstrate that the proposed algorithm achieves robust and accurate UAV tracking, maintaining a root-mean-square error (RMSE) within 1 meter and an average relative tracking error below 2.5% under low SNR conditions. These results highlight the method's potential for reliable UAV surveillance in challenging operational environments.

This toolkit is designed for prison governors and staff to improve the quality of food. It will help those involved in the preparation and production of food in women’s prisons to reflect on existing practices and to identify ways forward to improve the relationship between women and food in prison. The material in the toolkit is drawn from ESRC- and British Academy/Leverhulme-funded qualitative studies that focused on the role of food in women’s prisons. It has been devised in partnership with His Majesty’s Prison and Probation Service and charitable organisations including Women in Prison, Food Behind Bars, Food Matters and PACT.