Ludwig-Maximilians-Universität München
Open Access LMU ( Ludwig-Maximilians-Univ. München)Not a member yet
40914 research outputs found
Sort by
No substantial neurocognitive impact of COVID-19 across ages and disease severity
Get PDFBackground : Compared to intensive care unit patients with SARS-CoV-2 negative acute respiratory tract infections, patients with SARS-CoV-2 are supposed to develop more frequently and more severely neurologic sequelae. Delirium and subsequent neurocognitive deficits (NCD) have implications for patients’ morbidity and mortality. However, the extent of brain injury during acute COVID-19 and subsequent NCD still remain largely unexplored. Body-fluid biomarkers may offer valuable insights into the quantification of acute delirium, brain injury and may help to predict subsequent NCD following COVID-19.
Methods : In a multicenter, observational case-control study, conducted across four German University Hospitals, hospitalized adult and pediatric patients with an acute COVID-19 and SARS-CoV-2 negative controls presenting with acute respiratory tract infections were included. Study procedures comprised the assessment of pre-existing neurocognitive function, daily screening for delirium, neurological examination and blood sampling. Fourteen biomarkers indicative of neuroaxonal, glial, neurovascular injury and inflammation were analyzed. Neurocognitive functions were re-evaluated after three months.
Results : We enrolled 118 participants (90 adults, 28 children). The incidence of delirium [85 out of 90 patients (94.4%) were assessable for delirium) was comparable between patients with COVID-19 [16 out of 61 patients (26.2%)] and SARS-CoV-2 negative controls [8 out of 24 patients (33.3%); p > 0.05] across adults and children. No differences in outcomes as measured by the modified Rankin Scale, the Short-Blessed Test, the Informant Questionnaire on Cognitive Decline in the Elderly, and the pediatrics cerebral performance category scale were observed after three months. Levels of body-fluid biomarkers were generally elevated in both adult and pediatric cohorts, without significant differences between SARS-CoV-2 negative controls and COVID-19. In COVID-19 patients experiencing delirium, levels of GFAP and MMP-9 were significantly higher compared to those without delirium.
Conclusions : Delirium and subsequent NCD are not more frequent in COVID-19 as compared to SARS-CoV-2 negative patients with acute respiratory tract infections. Consistently, biomarker levels of brain injury indicated no differences between COVID-19 cases and SARS-CoV-2 negative controls. Our data suggest that delirium in COVID-19 does not distinctly trigger substantial and persistent subsequent NCD compared to patients with other acute respiratory tract infections
Pragmatic algorithm for visual assessment of 4-Repeat tauopathies in [18F]PI-2620 PET Scans
Get PDFAim : Standardized evaluation of [18F]PI-2620 tau-PET scans in 4R-tauopathies represents an unmet need in clinical practice. This study aims to investigate the effectiveness of visual evaluation of [18F]PI-2620 images for diagnosing 4R-tauopathies and to develop a straight-forward reading algorithm to improve objectivity and data reproducibility.
Methods : A total of 83 individuals with [18F]PI-2620 PET scans were included. Participants were classified as probable 4R-tauopathies (n = 29), Alzheimer's disease (AD) (n = 20), α-synucleinopathies (n = 15), and healthy controls (n = 19) based on clinical criteria. Visual assessment of tau-PET scans (choice: 4R-tauopathy, AD-tauopathy, no-tauopathy) was conducted using either 20–40-minute or 40–60-minute intervals, with raw (common) and cerebellar grey matter scaled standardized reading settings (intensity-scaled). Two readers evaluated scans independently and blinded, with a third reader providing consensus in case of discrepant primary evaluation. A regional analysis was performed using the cortex, basal ganglia, midbrain, and dentate nucleus. Sensitivity, specificity, and interrater agreement were calculated for all settings and compared against the visual reads of parametric images (0–60-minutes, distribution volume ratios, DVR).
Results : Patients with 4R-tauopathies in contrast to non-4R-tauopathies were detected at higher sensitivity in the 20–40-minute frame (common: 79%, scaled: 76%) compared to the 40–60-minute frame (common: 55%, scaled: 62%), albeit with reduced specificity in the common setting (20–40-min: 78%, 40–60-min: 95%), which was ameliorated in the intensity-scaled setting (20–40-min: 91%, 40–60-min: 96%). Combined assessment of multiple brain regions did not significantly improve diagnostic sensitivity, compared to assessing the basal ganglia alone (76% each). Evaluation of intensity-scaled parametric images resulted in higher sensitivity compared to intensity-scaled static scans (86% vs. 76%) at similar specificity (89% vs. 91%).
Conclusion : Visual reading of [18F]PI-2620 tau-PET scans demonstrated reliable detection of 4R-tauopathies, particularly when standardized processing methods and early imaging windows were employed. Parametric images should be preferred for visual assessment of 4R-tauopathies
DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback
No full textDefining a reward function is usually a challenging but critical task for the system designer in reinforcement learning, especially when specifying complex behaviors. Reinforcement learning from human feedback (RLHF) emerges as a promising approach to circumvent this. In RLHF, the agent typically learns a reward function by querying a human teacher using pairwise comparisons of trajectory segments. A key question in this domain is how to reduce the number of queries necessary to learn an informative reward function since asking a human teacher too many queries is impractical and costly. To tackle this question, we propose DUO, a novel method for diverse, uncertain, on-policy query generation and selection in RLHF. Our method produces queries that are (1) more relevant for policy training (via an on-policy criterion), (2) more informative (via a principled measure of epistemic uncertainty), and (3) diverse (via a clustering-based filter). Experimental results on a variety of locomotion and robotic manipulation tasks demonstrate that our method can outperform state-of-the-art RLHF methods given the same total budget of queries, while being robust to possibly irrational teachers
Ähnlichkeit
No full textÄhnlichkeit ist ein elementares Phänomen, das sich als Übereinstimmung von Objekten auf Basis gemeinsamer Merkmale manifestiert. Aufgrund ihrer Vielschichtigkeit ist das Denksystem der Ähnlichkeit eingebettet in ein komplexes Wechselspiel kulturell geprägter und individueller Wahrnehmungen, das immer wieder rekontextualisiert werden muss. Dieses Heft verbindet naturwissenschaftliche, geisteswissenschaftliche und künstlerische Perspektiven und zeigt, wie Ähnlichkeit algorithmisch operationalisiert wird und ähnlichkeitsstiftende Muster im digitalen Bild produktiv werden
Prolonged time to breast cancer surgery and the risk of metastasis: an explorative simulation analysis using epidemiological data from Germany and the USA
Get PDF
Data in Personalized Nutrition: Bridging Biomedical, Psycho-behavioral, and Food Environment Approaches for Population-wide Impact
Get PDFEditorial: ICF in teaching, training and education - retrospective for future concepts - what remains of 20 years of ICF in education?
Get PDFNon-relapse mortality with bispecific antibodies: A systematic review and meta-analysis in lymphoma and multiple myeloma
Get PDFBispecific antibodies (BsAb) are associated with distinct immune-related toxicities that impact morbidity and mortality. This systematic review and meta-analysis examined non-relapse mortality (NRM) with BsAb therapy in B-cell non-Hodgkin lymphoma (NHL) and multiple myeloma (MM). A PubMed and Embase search up to October 2024 identified 29 studies (21 NHL, 8 MM) involving 2,535 patients. The overall NRM point estimate was 4.7% (95% confidence interval [CI] 3.4%–6.4%), with a median follow-up of 12.0 months. We noted no significant difference in NRM across disease entities (NHL: 4.2%, MM: 6.2%, p = 0.22). In NHL, prespecified subgroup analyses revealed increased NRM in real-world studies compared to clinical trials. For MM, an association between NRM and higher response rates and longer follow-up was noted. Meta-regression comparing BsAb and CAR-T therapies (n = 8,592) showed no significant NRM difference when accounting for key study-level confounders (p = 0.96). Overall, infections were the leading cause of NRM, accounting for 71.8% of non-relapse deaths. Of the infection-related deaths, 48% were attributed to COVID-19. In a pre-specified sensitivity analysis excluding COVID-19 fatalities, the overall NRM estimate was 3.5% (95% CI 2.6%–4.6%). Taken together, these results provide a benchmark for the estimated NRM with BsAb therapy and highlight the paramount importance of infection reporting, prevention, and mitigation