Hong Kong University of Science and Technology
Hong Kong University of Science and Technology Institutional RepositoryNot a member yet
162821 research outputs found
Sort by
Linking Transition Metal Concentration and Oxidative Potential of PM<sub>2.5</sub> by Acellular Assays: Reaction Mechanisms and Metal-Organic Mixing Effects
No full text
Tuning the Optoelectronic and Chiroptical Properties of Colloidal Nanocrystals for Quantum Information Science
No full textVariance-Reduced First-Order Methods for Deterministically Constrained Stochastic Nonconvex Optimization with Strong Convergence Guarantees
No full textIn this paper, we study a class of deterministically constrained stochastic nonconvex optimization problems. Existing methods typically aim to find an ϵ-expectedly feasible stochastic stationary point, where the expected violations of both constraints and first-order stationarity are within a prescribed tolerance ϵ. However, in many practical applications, it is crucial that the constraints be nearly satisfied with certainty, making such an ϵ-stochastic stationary point potentially undesirable due to the risk of substantial constraint violations. To address this issue, we propose single-loop variance-reduced stochastic first-order methods, where the stochastic gradient of the stochastic component is computed using either a truncated recursive momentum scheme or a truncated Polyak momentum scheme for variance reduction, while the gradient of the deterministic component is computed exactly. Under the error bound condition with a parameter θ ≥1 and other suitable assumptions, we establish that these methods respectively achieve sample complexity and first-order oracle complexity of (Formula presented) for finding an ϵ-surely feasible stochastic stationary point (formula presented) with logarithmic factors hidden), where the constraint violation is within ϵ with certainty, and the expected violation of first-order stationarity is within ϵ. For θ =1, these complexities reduce to (formula presented), respectively, which match, up to a logarithmic factor, the best-known complexities achieved by existing methods for finding an ϵ -stochastic stationary point of unconstrained smooth stochastic nonconvex optimization problems.</p
Impact of different preferred functional outcomes on the results of the pre-hospital antifibrinolytics for traumatic coagulopathy and hemorrhage (PATCH-trauma) trial
No full textRationale: Ordinal endpoints offer more detailed outcome assessment than mortality, though some disabilities may be considered worse than death by patients. Objectives: To evaluate how different Glasgow Outcome Scale - Extended (GOS-E) hierarchical structures affect effect estimation in the PATCH-Trauma trial. Methods: This post-hoc analysis of the PATCH-Trauma trial compared pre-hospital tranexamic acid (TXA) to placebo in severely injured patients at risk of acute traumatic coagulopathy. The study included patients with complete 6-month outcomes. Using generalized pairwise comparisons, researchers analyzed the 8-point GOS-E, ICU length-of-stay, and thromboembolic complications. Results were reported using win ratio (WR) statistics, with values above one favoring TXA. The analysis explored GOS-E with mortality rankings adjusted based on different outcomes being considered worse than death. Main Results: The study included 1107 patients (546 placebo, 561 TXA). Original GOS-E results showed a neutral WR (1.03; 95 % CI 0.90–1.19). When death was considered preferable to progressively less impaired functionality (from vegetative state to upper moderate disability), the WR decreased to 0.86 (95 % CI 0.75–1.00). Conclusion: The assessment of TXA therapy's efficacy and safety may vary depending on scenarios where certain functional outcomes are considered worse than death. This analytical approach could inform functional outcome assessment in future acute care trials.</p
Bio2Vol: Adapting 2D Biomedical Foundation Models for Volumetric Medical Image Segmentation
No full text2D biomedical foundation models (FM) have demonstrated remarkable capabilities in 2D medical image segmentation across various modalities, with text-prompted approaches offering scalable analysis that facilitate integration with LLMs and clinical application. Adapting these models for 3D medical image segmentation can leverage their rich visual features while enabling text-prompted volumetric image segmentation. However, efficient adaptation poses significant challenges due to the substantial disparity between 2D and 3D medical images and the necessity to establish text-volume alignment. To address these limitations, we propose Bio2Vol, a novel adaptation framework that enables text-prompted 2D biomedical FMs to effectively handle volumetric data. Specifically, (1) To bridge the dimensional disparity, we propose a Dual-Rate Sampling strategy (DRS) that processes inter slices within a volume at both sparse and dense intervals, capturing global contexts and local details; (2) To enhance volumetric feature representation, a Cross-slice Dual-head Attention (CSDHA) is built upon the intra-slice features by repurposing existing pre-trained attention modules for parameter-efficient inter-slice information fusion; and (3) To establish text-volume understanding, a Semantic Text-Visual Alignment loss (SAT) is used to extend the existing 2D text-visual alignment to the volumetric domain. Using BiomedParse as a demonstration case, extensive evaluation across 11 medical datasets across diverse anatomical regions and modalities shows that Bio2Vol significantly improves 3D medical image segmentation performance, enhancing DSC by 4.72% on Amos22 dataset with substantial improvements across MSD tasks. Code will be available https://github.com/JiaxinZhuang/Bio2Vol.</p
Making waves: Conductive materials in anaerobic digestion: A sustainable pathway or a hidden carbon burden?
No full textConductive material-mediated anaerobic digestion (AD) systems offer a promising solution to enhance methane production, yet its sustainability and economic viability require holistic evaluation. In this study, we systematically assess the typical conductive materials, namely, biochar, iron-based material, and biochar-iron composites, through integrated life cycle and cost-benefit analyses of 219 experimental cases. Biochar-iron composites achieved the highest methane yield improvement (36 %), while iron-based materials posed significant carbon burdens (contributing up to 44 % of system emissions). Crucially, material recycling (five cycles) reduced iron's carbon footprint by 72 %, and digestate valorization into biochar further lowered net emissions by 113.8–184.9 %. Economically, iron-based materials outperformed biochar in profitability (220 USD/ton volatile solids), and combining material recovery with digestate valorization boosted net profits by 191.8–264.8 %. The findings demonstrate that prioritizing biochar-iron composites for performance and iron-based materials with recovery for cost-effectiveness, alongside closed-loop design, can reconcile environmental and economic goals. This work provides actionable pathways to optimize conductive material-enhanced AD systems for scalable, sustainable waste-to-energy conversion.</p
Graphitic biochar-anammox achieved by multi-heme-based extracellular electron transfer
No full textThe anammox process decarbonizes nitrogen removal by avoiding greenhouse gas emissions and organic carbon demand, yet its reliance on nitrification and denitrification undermines these advantages. To address this, we developed a biochar-assisted anammox system that leverages graphitic defects as redox-active sites to enable interspecies, multi-heme extracellular electron transfer (EET). Biochar produced at 800 °C for 4 h (BC800–4 h) exhibited the greatest graphitic defect density and the highest electron accepting capacity, uniquely exceeding the daily stoichiometric electron demand for complete ammonium oxidation in the present study. Metagenomic and in vitro assays revealed that BC800–4 h promoted hydroxylamine-dependent ammonium oxidation by anaerobic ammonia-oxidizing bacteria (AnAOB) via EET. A cooperative microbial network was identified: AnAOB in suspension supplied heme precursors, while ammonia-oxidizing and denitrifying bacteria colonizing the biochar facilitated heme assembly and transport. This partitioning enabled direct electron transfer to biochar, achieving 62 % nitrogen removal without exogenous nitrite and reducing N₂O emissions by 28 %. The pore-size-dependent reduction in graphitic defects suggests that large molecular-weight biological channels (>10 kDa) are essential for electron transfer between anammox consortia and biochar. Our findings indicate an opportunity to develop a biochar-anammox reactor—with suspended AnAOB and a fixed-bed biochar biofilm—to exploit this synergy for efficient and low-emission nitrogen removal.</p
TSAR: A two-stage approach to motion artifact reduction in OCTA images
No full textOptical Coherence Tomography Angiography (OCTA) is an innovative and non-invasive imaging technique that leverages motion contrast imaging to generate angiographic images from high-resolution volumetric blood flow data rapidly. However, OCTA imaging is vulnerable to various artifacts induced by eye movements, including displacement artifacts, duplicated scanning artifacts, and white line artifacts. Previous methods that attempted to mitigate eye motion artifacts necessitated costly hardware upgrades. However, despite the availability of advanced eye-tracking hardware and software correction in commercial machines, motion artifacts persist in real-world usage. Recently developed cost-effective learning-based methods only focus on the removal of white line artifacts while neglecting the displacement artifacts and duplicated scanning artifacts. To address this challenge, we propose a comprehensive framework, TSAR, to remove three types of eye motion artifacts in OCTA images. In the first stage, we leverage the intrinsic axial and directional attributes of these artifacts in the first phase to develop an innovative hierarchical transformer network. This network is designed to capture global-wise, local-wise, and vertical-wise features effectively while also removing displacement and duplicate scanning artifacts. Afterward, we leverage the contextual information and develop a residual conditional diffusion model (RCDM) to remove the white line artifacts. By applying our TSAR to the degraded OCTA images, we aim to eliminate all three types of motion artifacts. We evaluate the superior performance of our proposed methodology in artifact removal and image quality enhancement compared to other methods by conducting experiments on both synthetic and real-world OCTA images. The code is available at https://github.com/btma48/TSAR</p