Opin vísindi

 

Nýlega bætt við

Verk
Advancing Land-Cover and Land-Use Classification with Semantically Consistent and Data-Efficient Deep Learning on Multi-Source Earth Observation Data
(University of Iceland, School of Engineering and Natural Sciences, Faculty of Industrial Engineering, Mechanical Engineering and Computer Science, 2026-10) Sharma, Surbhi; Prof. Dr. Morris Riedel; Faculty of Industrial Engineering, Mechanical Engineering and Computer Science (UI); Iðnaðarverkfræði-, vélaverkfræði- og tölvunarfræðideild (HÍ); School of Engineering and Natural Sciences (UI); Verkfræði- og náttúruvísindasvið (HÍ)
Land-use and Land-cover (LULC) mapping from Earth Observation (EO) imagery plays a fundamental role in environmental monitoring, urban planning, agricultural management, and climate change assessment. Recent advances in satellite missions, Machine Learning (ML), Deep Learning (DL), and foundation models have significantly improved large-scale mapping capabilities. However, important challenges remain, including semantic inconsistencies between satellite imagery and in-situ observations, the lack of large-scale benchmark datasets with detailed LULC information for training and evaluating advanced DL models, heterogeneous data sources, mixed pixel effects, limited availability of annotated training data, and the need for computationally scalable processing frameworks. Addressing these challenges requires methodologies that integrate multiple data modalities while improving semantic consistency, data efficiency, and spatial representation. This thesis presents a comprehensive framework for advancing satellite-based LULC mapping through four complementary research directions. First, it investigates the semantic relationship between satellite observations and in-situ field data by leveraging geo-tagged field photographs as an intermediate observational layer. The analysis demonstrates that apparent disagreements between satellite-derived and field-based information often arise not only from limitations of satellite observations but also from ambiguities in field interpretation and inconsistencies in land-cover (LC) taxonomies. Second, the thesis produces the Sen4Map, a large-scale benchmark dataset, integrating Sentinel-2 imagery with harmonized Land Use/Cover Area Frame Survey (LUCAS) field observations and ancillary information. The benchmark provides semantically rich annotations designed to support the development and evaluation of robust DL models for LULC mapping. Third, the thesis develops methods for sub-pixel LULC estimation to capture fine-scale landscape heterogeneity beyond the native spatial resolution of satellite imagery. As a proof of concept, the proposed framework estimates built-up fractions from Sentinel-2 imagery using geospatial foundation model embeddings, demonstrating the potential of pretrained spectral-spatial representations for sub-pixel characterization. And finally, the thesis explores data-efficient learning strategies for EO by investigating few-shot learning with meta-learning on publicly available Remote Sensing (RS) benchmark datasets. These approaches demonstrate improved generalization under limited annotation availability, reducing the dependence on extensive labeled datasets while maintaining competitive mapping performance across diverse environments. Beyond methodological advances, the thesis also contributes scalable computational iii solutions through a Hierarchical Data Format, version 5 (HDF5)-based data architecture optimized for High-Performance Computing (HPC) environments, enabling efficient storage, access, and processing of large multi-source EO datasets. Collectively, the proposed methodologies contribute toward more semantically consistent, data-efficient, and scalable satellite-based LULC mapping systems. As EO transitions toward foundation models, unified embedding representations, and agentic geospatial intelligence, the contributions presented in this thesis establish the datasets, semantic frameworks, and embedding-centric methodologies required to support the next generation of LULC mapping.
Verk
Evaluating the Validity of PHQ-9 Scores: A Multi-Method Psychometric Investigation of Sources of Variance Beyond Depression Severity
(University of Iceland, School of Health Sciences, Faculty of Psychology, 2026-10-07) Kristófersdóttir, Kristín Hulda; Fanney Þórsdóttir; Faculty of Psychology (UI); Sálfræðideild (HÍ); School of Health Sciences (UI); Heilbrigðisvísindasvið (HÍ)
The Patient Health Questionnaire-9 (PHQ-9) is one of the most widely used self-report measures of depressive symptoms in clinical practice and research. Its scores are often interpreted as indicators of depression severity, yet this interpretation depends on the extent to which responses reflect depressive symptomatology rather than other influences. The aim of this thesis was to evaluate the validity of PHQ-9 scores by examining sources of variance beyond depression severity that may influence item responses. The thesis comprised four interrelated studies conducted with Icelandic samples. Paper I examined the psychometric properties of an existing Icelandic translation of the PHQ-9 using Mokken scale analysis (MSA) and used cognitive interviews to investigate possible sources of weak item functioning. Paper II introduced a revised Icelandic translation of the PHQ-9 and evaluated its psychometric properties using MSA. Paper III used the method of paired comparisons to examine whether PHQ-9 items differ in their perceived sensitivity, that is, the extent to which respondents experience them as uncomfortable to disclose. Paper IV modeled socially desirable responding as a measured cause of method variance in the PHQ-9 to examine whether it systematically influences item responses and whether that influence varies with item sensitivity. The findings showed that the original Icelandic PHQ-9 had suboptimal scalability and that this was associated with several problems, including translation errors, ambiguous wording, multi-barreled item content, symptom overlap, and other response-process issues. The revised Icelandic translation performed better psychometrically, with stronger scalability and a more coherent ordering of respondents along a depression severity continuum. However, not all problems identified in the original version were resolved. The results further showed that PHQ-9 items differ systematically in perceived sensitivity. Items typically grouped as cognitive/affective were generally judged more uncomfortable to disclose than items grouped as somatic. Finally, socially desirable responding was a source of method variance in PHQ-9 responses, systematically lowering symptom reporting while accounting for considerably less variance than depression severity. This method effect varied across items but did not correspond to their perceived sensitivity. In sum, the findings suggest that PHQ-9 scores may reflect more than depression severity alone. The results support a cautious and context-sensitive interpretation of PHQ-9 sum scores. The PHQ-9 may still be useful as a brief indicator of depressive symptoms, particularly in its revised Icelandic form, but its scores should not be treated as fully transparent or pure measures of depression severity. The thesis underscores the importance of examining how instrument design, item content, translation, and response processes shape the meaning of psychological measurement scores.
Verk
Fabrication and characterization of Si1−xGex nanocrystals in as-grown and annealed structures: a comparative study
(2019-09-17) Sultan, Muhammad Taha; Maraloiu, Adrian Valentin; Stavarache, Ionel; Gudmundsson, Jón Tómas; Manolescu, Andrei; Teodorescu, Valentin Serban; Ciurea, Magdalena Lidia; Svavarsson, Halldór Gudfinnur; Verkfræðideild
Multilayer structures comprising of SiO2/SiGe/SiO2 and containing SiGe nanoparticles were obtained by depositing SiO2 layers using reactive direct current magnetron sputtering (dcMS), whereas, Si and Ge were co-sputtered using dcMS and high-power impulse magnetron sputtering (HiPIMS). The as-grown structures subsequently underwent rapid thermal annealing (550-900 degrees C for 1 min) in N-2 ambient atmosphere. The structures were investigated using X-ray diffraction, high-resolution transmission electron microscopy together with spectral photocurrent measurements, to explore structural changes and corresponding properties. It is observed that the employment of HiPIMS facilitates the formation of SiGe nanoparticles (2.1 +/- 0.8 nm) in the as-grown structure, and that presence of such nanoparticles acts as a seed for heterogeneous nucleation, which upon annealing results in the periodically arranged columnar self-assembly of SiGe core-shell nanocrystals. An increase in photocurrent intensity by more than an order of magnitude was achieved by annealing. Furthermore, a detailed discussion is provided on strain development within the structures, the consequential interface characteristics and its effect on the photocurrent spectra.
Verk
Structural brain alterations following adult non-CNS cancers : a systematic review of the neuroimaging literature
(2019-05-04) Amidi, Ali; Wu, Lisa M.; Department of Psychology
Background: Cancer and cancer treatments may impact the brain through several pathways leading to cognitive impairment. Neuroimaging evidence has begun to elucidate the neurobiological underpinnings of cancer-related cognitive impairment. The aim of this paper was to systematically review available literature on structural brain alterations following adult non-central nervous system (CNS) cancers and associated treatments. Methods: This review followed PRISMA guidelines and was registered in PROSPERO (ID#107387). Comprehensive searches were conducted in June 2018 using PubMed and Web of Science. Inclusion criteria were English peer-reviewed journal articles of formal, controlled studies that examined structural neuroimaging outcomes in adult non-CNS cancer patients and survivors. Selected articles were assessed for quality and risk of bias using the National Institutes of Health Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. Results: Thirty-six publications of prospective and cross-sectional studies met inclusion criteria and were included. Structural brain alterations following cancer and its treatment were reported in a majority of the publications as evidenced by reduced global and local gray matter volumes, impaired white matter microstructural integrity, and brain network alterations. Structural alterations were most often evident when cancer-treated groups were compared with healthy controls, and more subtle when compared with cancer controls. Regarding the existence of pretreatment impairments, the evidence was equivocal. There was significant between-study heterogeneity in imaging analytical approaches and use of statistical adjustments. Over half reported associations with cognitive outcomes, though regions and associated cognitive domains were heterogeneous. Conclusions: Structural brain alterations following cancer and cancer treatments were reported in a majority of the reviewed studies. However, the extent of observed alterations depended on the choice of comparison groups. Methodological issues exist that will need to be addressed systematically to ensure the validity of findings. Large-scale prospective studies with extended assessment points are warranted to replicate and build upon initial findings.
Verk
Long-term cognitive dysfunction after radiation therapy for primary brain tumors
(2019-05-04) Haldbo-Classen, Lene; Amidi, Ali; Wu, Lisa M.; Lukacova, Slavka; Oettingen, Gorm von; Gottrup, Hanne; Zachariae, Robert; Høyer, Morten; Department of Psychology
Background: The extent of radiation therapy (RT)-induced changes in cognitive function is unknown. RT with protons instead of photons spares the healthy brain tissue more and is believed to reduce the risk of cognitive dysfunction. There is modest knowledge on which parts of the brain we need to spare, to prevent cognitive dysfunction. To uncover which cognitive domains is most affected, we compared cognitive functioning in brain tumor patients treated with neurosurgery and RT with brain tumor patients treated with neurosurgery alone. Methods: A cross-sectional study assessing cognitive function in 110 patients with a primary brain tumor grades I–III or medulloblastoma (grade IV) treated at Aarhus University Hospital (AUH), Denmark between 2006 and 2016. Two cohorts were established: a cohort of 81 brain tumor patients who had received neurosurgery followed by RT (RT+), and a cohort of 29 brain tumor patients who had only received neurosurgery (RT–). The patients underwent questionnaires and neuropsychological assessment with standardized tests. Results: Mean age was 53.5 years with an average time since diagnosis of 7.3 years. Compared with normative data, lower average scores were observed for the entire group on domains concerning of verbal learning and memory (p <.001), attention and working memory (p <.001), processing speed (p <.001), and executive functioning (p <.001). Compared to RT– patients, RT + patients scored lower on domains concerning processing speed (p =.04) and executive function (p =.05) and had higher impairment frequency on verbal fluency (p =.02) with 16% of patients exceeding 1.5 SD below normative data. Conclusions: Our results indicate that treatment, including RT, for a primary brain tumor may have negative long-term impact on cognitive function, especially on processing speed and executive function.

Flokkar í Opnum vísindum

Veldu flokk til að skoða.

Niðurstöður 1 - 9 af 9

Öll gögn í Opnum vísindum eru vernduð af ákvæðum höfundalaga og með öllum réttindum áskildum, nema annað sé tekið fram.