154092 research outputs found
Sort by
Deep learning-based vegetation canopy height mapping with polarimetric SAR:Application of a Polarization Fusion U-Net in Gabon’s tropical forests
Get PDFForests provide essential ecosystem services, including carbon storage and the conservation of biodiversity. This highlights the need for accurate and scalable methods to assess forest structure. Active remote sensing, particularly Synthetic Aperture Radar (SAR), has significant potential for estimating forest structure owing to its ability to penetrate vegetation layers and interact with different forest elements. In particular, L- and P-band SAR signals are able to penetrate canopy layers, making them suitable for retrieving forest structure information.We present a novel approach utilizing full-polarimetric SAR backscatter data to estimate canopy height as relevant forest structure variable. To capture the non-linear complex relationship between the SAR data and the vegetation canopy height, we propose a Polarization Fusion U-Net (PF-Unet) designed to enhance canopy height estimation from SAR backscatter data by effectively utilizing multi-polarization channels (e.g., HH, HV, and VV). Specifically, the proposed model is tailored to the physical properties of the SAR backscatter, incorporating: (a) a polarization fusion layer, (b) attention gates applied to each layer in the decoder blocks, and (c) Exponential Linear Unit (ELU) activation and Huber loss functions. To assess the potential of the PF-Unet model, L- and P-band SAR data, collected over the tropical forests of Gabon, were separately used. The model was evaluated in a complex tropical forest environment and compared against traditional Machine Learning (ML) approaches (Random Forest (RF) and Light Gradient Boosting Machine (LGBM)) as well as the standard U-Net model. The PF-Unet consistently outperformed all the baselines for both SAR datasets. The PF-Unet model achieved an RMSE of 4.35 m (15.73%) for L-band and 4.43 m (15.95%) for P-band, which was an improvement over the U-Net model’s RMSE of 5.02 m (18.15%) and 4.59 m (16.53%), respectively. This suggests promising potential for enhanced canopy height estimation, which is particularly valuable for upcoming spaceborne missions like NISAR and BIOMASS
Breast cancer prediction using mammography exams for real hospital settings
Get PDFBreast cancer prediction models for mammography assume that annotations are available for individual images or regions of interest (ROIs), and that there is a fixed number of images per patient. These assumptions do not hold in real hospital settings, where clinicians provide only a final diagnosis for the entire mammography exam (case). Since data in real hospital settings scales with continuous patient intake, while manual annotation efforts do not, we develop a framework for case-level breast cancer prediction that does not require any manual annotation and can be trained with case labels readily available at the hospital. Specifically, we propose a two-level multi-instance learning (MIL) approach at patch and image level for case-level breast cancer prediction and evaluate it on two public and one private dataset. We propose a novel domain-specific MIL pooling observing that breast cancer may or may not occur in both sides, while images of both breasts are taken as a precaution during mammography. We propose a dynamic training procedure for training our MIL framework on a variable number of images per case. We show that our two-level MIL model can be applied in real hospital settings where only case labels, and a variable number of images per case are available, without any loss in performance compared to models trained on image labels. Only trained with weak (case-level) labels, it has the capability to point out in which breast side, mammography view and view region the abnormality lies.</p
Acquisition of respiratory surface EMG:A systematic literature review of electrode configurations and methodological reporting
Get PDFBackground: Mechanical ventilation provides life-saving support to patients with respiratory failure, but inadequately tailored settings can lead to respiratory muscle dysfunction and poor patient outcomes. Surface electromyography (sEMG) offers a non-invasive modality to monitor respiratory muscle function. However, variability in acquisition setups limits the comparability of study findings and hinders broad clinical implementation. Therefore, we systematically appraised setup rationales and reporting quality in respiratory sEMG literature.Methods: The MEDLINE ALL, Embase, and Web of Science databases were systematically searched on 19 September 2024 for studies reporting original respiratory sEMG data in adults during spontaneous breathing. sEMG methodology was extracted in accordance with the reporting guidelines of the International Society of Electrophysiology and Kinesiology and analyzed by target muscle and medical domain.Results: 240 out of 402 unique articles were included. The diaphragm was the most studied respiratory muscle (61%) with 48 unique setups out of 160 descriptions. Diaphragm setups with small inter-electrode distances (IEDs) were most common (n = 138, 86%). Large IED setups were predominantly applied in ICU (n = 8, 36%) and COPD (n = 5, 23%) populations. Setups for non-diaphragmatic respiratory muscles typically featured one or two dominant positions grounded in methodological studies. Reporting quality was low with a median of 5 out of 10 recommended items documented.Conclusion: This review reveals substantial diversity of diaphragm sEMG setups, reflecting differences in clinical contexts and study populations. The setups for extra-diaphragmatic muscles were more consistent and methodologically grounded. Muscle- and context-specific guidelines are essential to improve consistency and support clinical implementation of respiratory sEMG.</p
Experimenting with Reaction Systems using Graph Transformation and GROOVE
No full textWe explore the capabilities of GROOVE, a state-of-the-art toolset based on graph transformation systems, to perform different kinds of analyses of Reaction Systems, ranging from reachability and causal analysis to model checking. Our results are encouraging, as in the presence of large state spaces GROOVE improves the time required for both reachability and causal analyses by an order of magnitude, compared to other available tools. From the point of view of GROOVE, the implementation of Reaction Systems provided some interesting insights on the most convenient way to model certain computational requirements through negative and nested application condition
Assessing long-term temporal changes in myocardial infarction treatment and outcomes:Chances and challenges
Get PDFSuspended Z-cut lithium niobate waveguides for stimulated Brillouin scattering
Get PDFOn-chip stimulated Brillouin scattering (SBS) has recently been demonstrated in thin-film lithium niobate (TFLN), an emerging material platform for integrated photonics offering large electro-optic and nonlinear properties. While previous works on SBS in TFLN have focused on surface SBS, in this contribution we experimentally demonstrate, for the first time, backward intra-modal SBS generation in suspended Z-cut TFLN waveguides. Our results show trapping of multiple acoustic modes in this structure, featuring a multi-peak Brillouin gain spectrum due to the excitation of higher-order acoustic modes. The findings expand the TFLN waveguide platform exploration for SBS interactions and provide a crucial step toward realizing advanced optical processors for sensors or microwave signals integrated on TFLN.</p
Balancing acts
Get PDFGrant Kajii and Polak (2000, Journal of Economic Theory) have identified weak decomposability as the key axiom to characterise the betweenness class for acts without the assumption of probabilistic sophistication, leaving, however, a small gap between necessary and sufficient conditions. We show how the balancing perspective naturally leads to a full characterisation, for monetary acts on a finite state space. In our first theorem, we avoid some technical issues by working with the implicit notion of soundness, reflecting the proper working of a sliding balance. We sharpen the result using ratio monotonicity as the `missing' axiom, already known for lotteries as a consequence of first order stochastic dominance, but not necessarily satisfied under weak decomposability for acts. It guarantees soundness, and in fact induces a semi-strict form of monotonicity of utility in balance setpoints. Finally, we show how the balancing perspective offers an intuitive way to incorporate partial acts in the framework without reference to conditioning, replacing the notion of a null state by a nil outcome. This leads to a very straightforward betweenness axiom: the balance point of an act should be in between those of its parts. Our final theorem characterises the betweenness class for acts using this axiom of balance point betweenness, combined with a perhaps unexpected requirement of embeddability
Transdisciplinary perspectives on waiting for healthcare services in South Africa
Get PDFProviding all individuals with quality healthcare is a global priority. In South Africa, there is an accelerated need for healthcare services, particularly those provided in hospitals.However, resources to provide quality healthcare services in hospitals are limited. Additionally, natural and artificial variability of both healthcare demand patterns and resource availability leads to delays in South African hospitals. These delays can be addressed by adopting techniques in the discipline of Operations Research and Management Science (ORMS), which has been proven to drive efficiencies. Adopting these techniques in South African hospitals can successfully address the challenges around delays caused by variable demand patterns and the availability of resources. However, to successfully implement such interventions in South African hospitals, research calls for the consideration of several fundamental factors, such as the unique culture and context of the healthcare setting.This research responds to this need by providing a clear understanding of the challenges and opportunities that the application of the ORMS discipline in South African hospitals poses. Our research further contributes towards the call for a more resilient and responsive health system in South Africa.In this research, we took a transdisciplinary approach to identify these unique challenges and opportunities by considering the disciplines of ORMS and education and training. With the hospital setting as a health system with multiple stakeholders, we approached the research from a patient, clinical staff and hospital organisational perspective.From the patient perspective of delays in healthcare settings, we find that waiting is not always detrimental and that there is a need to approach waiting time in healthcare settings from a nuanced approach in ORMS research. To this end, we provide definitions for the terms access time and waiting time that are frequently used to refer to delays in healthcare settings. We also highlight that the quality of patient care and quality of service healthcare are two distinguishable characteristics since the quality of patient care takes voluntary and involuntary waiting time into account. Juxtaposing the quality of patient care and quality of service is, therefore, a prerequisite in solving the waiting time problem in healthcare services as it addresses the variability between the supply of capacity and the demand for healthcare services.The clinical staff perspective of our research provides insights into the important factors associated with the mental health of nurses in South Africa. Specifically, we propagate that researchers and practitioners of ORMS in healthcare take cognisance of the complexities associated with feelings of burnout and emotional exhaustion of nurses
The problem of face image morphing in identification documents:Analysis, prevention and detection
Get PDFThis dissertation focuses on the problem of face image morphing in identification documents. Given the face images of two (or more) individuals, we can create a composite face image (a morph) using computer software that looks like both of them. One of the contributors (the accomplice) could use such a morph to apply for a new identification document, e.g. a passport. If that passport gets issued, the other contributor (the criminal) can then use it to conduct illegal activities and/or evade justice. This identity sharing scheme is known as a morphing attack and is a serious security vulnerability. In the first chapter, I introduce the subject of face morphing, talk about its historical use cases and analyse its causes and consequences. The second chapter is dedicated to improving morph generation algorithms, since being better prepared for what we might encounter in the real world is crucial. There, I present some improvements to a specific automated morph generation process, namely landmark-based morphing. The third chapter consists of a novel morphing prevention approach in ID document enrolment processes that, if implemented, would seriously limit the morphing attack threat. Next, I focus on morphing attack detection and present two methods based on image forensics that can be used to visualise morphing-related manipulation traces on digital images. Visualisation of morphing traces is very important as it can be used as evidence in a court of law. Finally, since the scale of the problem requires automated solutions, I use one of the aforementioned methods to create an automated morphing attack detection algorithm, so that any incriminating decision by the algorithm can be reviewed and explained by applying the corresponding image forensics method and visualising any possible traces
Photophysical Characterization of Layered Perovskites for Solar Cell Applications
No full textThe transition to renewable energy is crucial in combating global warming, with solar power offering a key solution. While silicon-based solar cells dominate the market, ongoing research explores new materials that could enhance efficiency and reduce costs.Among these materials, quasi-two-dimensional (quasi-2D) metal halide perovskites have emerged as promising candidates due to their impressive optical and optoelectronic properties. Compared to their three-dimensional (3D) counterparts, quasi-2D perovskites offer improved stability. In quasi-2D metal halide perovskite (MHP) films, large organic cations, known as spacer molecules, separate the inorganic layers, forming domains of varying thicknesses, referred to as n-phases. The coexistence of multiple n-phases within a single quasi-2D film leads to complex photoinduced exciton and charge carrier dynamics, which directly impact the solar cell performance.In photovoltaic (PV) devices, maximizing efficiency depends on the rapid extraction of photoexcited charge carriers before they recombine through either radiative (photon-emitting) or non-radiative (trap-assisted) pathways. However, in quasi-2D perovskites, the presence of different n-phases introduces additional complexity. Each domain exhibits a distinct bandgap, affecting how excitons (bound electron-hole pairs) and free charge carriers behave. Factors such as crystal composition, structure, orientation, and defects significantly influence these photophysical processes. Despite the potential of quasi-2D perovskites, existing photophysical models struggle to fully explain their exciton and charge carrier dynamics, and interpretations often remain inconsistent.This research aims to deepen our understanding of photoinduced exciton and charge carrier dynamics in quasi-2D perovskites. By investigating hot carrier thermalization, the directionality and time scales of transfer processes between low-n and high-n phases, and whether these involve excitons or charge carriers, this work aims to pave the way for optimizing quasi-2D perovskites for PV applications. Ultimately, these insights will contribute to the development of high-efficiency and stable PV technologies.<br/