Publikationer från Uppsala Universitet
Not a member yet
127603 research outputs found
Sort by
Challenges of using Artificial Intelligence in Credit Decision-Making
No full textArtificial Intelligence (AI) is transforming the financial sector, creating a major impact in credit decision making. Adopting machine learning tools in credit assessment has opportunities and challenges at the same time. While AI provides faster, accurate and more data-driven credit decisions, credit analysts on the other hand face some major challenges. This study explores the challenges that credit analysts face through the lens of institutional theory. After conducting semi-structured interviews with the credit analysts who are currently working in banks and financial institutions, this study finds some profound insights. Thematic analysis of this study reveals that transparency is a major challenge, as many credit analysts find it hard to interpret AI decisions to clients. Regulatory challenges add additional pressure for organizations with evolving AI frameworks like GDPR. Furthermore, cultural resistance like job displacement fears and the digital skills gap among professionals creates major concern. Despite all of these challenges, some credit analysts think of AI as a support tool rather than a replacement. They think that AI can increase their efficiency, accuracy, and effectiveness in credit assessment, though the ultimate decisions are made by them. It also provides insights for financial institutions who are looking for transparent, compliant, and culturally sensitive AI integration. Finally, this study applies institutional theory and its three dimensions, such as coercive, mimetic, and normative pressures, to find out how internal norms and external pressures shape AI implementation. Ultimately, this study contributes to both the theory and practice of AI adoption by understanding socio technical challenges
A Lab-on-a-silicon Chip Platform for Profiling Bacterial Metabolism
No full textRapid and reliable phenotypic antibiotic susceptibility testing (AST) is essential to address the growing challenge of antimicrobial resistance. Traditional AST methods rely on detecting bacterial growth, resulting in a long sample-to-result time. This thesis addresses this limitation by developing a lab-on-a-silicon-chip (LOSC) platform that directly collects and concentrates bacteria from clinical samples, following electrically monitoring bacterial metabolism in real time, offering a rapid, label-free approach for phenotypic AST. The work begins with the development of an integration process for combining nanoscale ion-sensitive field-effect transistor (ISFET) sensors with microfluidics, including the implementation of on-chip AgCl/Ag pseudo-reference electrodes (p-REs) for ISFET operation. These p-REs are fabricated using well-established semiconductor processes and are protected with a photoresist layer to enhance their stability. Their proximity to the ISFETs minimizes streaming potential effects under varying flow rates. To further improve measurement reliability, an AC gate biasing technique is employed to suppress signal drift, enabling stable, long-term real-time monitoring. Building on this foundation, a high-throughput LOSC platform with picoliter-scale microchambers is developed to collect and concentrate bacterial cells for AST. These chambers efficiently trap cells from dilute E. coli cultures, achieving local densities around 3×108 cells/mL – sufficiently high to induce detectable pH changes from cellular metabolism. Using this setup, the LOSC platform enables rapid AST of clinical E. coli isolates with a sample-to-result time below 20 minutes. In addition to AST, this platform provides insights into bacterial metabolic responses under antibiotic exposure. We have found that the changed acidification during the first hour of ampicillin treatment is induced by altered cellular metabolism rather than cell lysis. Liquid chromatography–mass spectrometry (LC-MS) analysis identifies acetate, succinate, and pyruvate as the primary organic acids that are responsible for pH reduction. Further experiments using ΔnuoG and ΔatpA mutants reveal that disruptions in energy metabolism alter acidification dynamics and antibiotic susceptibility. ΔnuoG mutants enhance malate and fumarate secretion via the glyoxylate shunt, while ΔatpA exhibits elevated pyruvate excretion and reduced sensitivity to ampicillin, suggesting a link between the tricarboxylic acid cycle (TCA) activity and bactericidal efficacy. To improve signal robustness for point-of-care applications, a lateral bipolar junction transistor (LBJT) is integrated with a nanoribbon field-effect transistor (NRFET) sensor as an internal amplifier. This LBJT-NRFET configuration achieves a current gain of over 20 and a 3–7× improvement in signal-to-noise ratio. The LBJT-NRFET maintains high pH sensitivity and exhibits strong immunity to external interference. Moreover, the fabrication process remains fully compatible with standard semiconductor fabrication technology, supporting scalable sensor integration. In summary, this work establishes a robust and scalable LOSC platform for rapid, phenotypic AST based on bacterial metabolism. Its fully electrical readout, semiconductor process compatibility, and high sensitivity make it well-suited for future applications in clinical diagnostics and microbial physiology studies
Early Career Progression in Young Adults With Coeliac Disease-A Register-Based Retrospective Cohort Study
No full textAim: To investigate early career progression and national insurance use in young adults with paediatric coeliac disease. Methods: We performed a register study of a population born in Sweden between 1976 and 1992. Coeliac disease was diagnosed before 15 years of age. The comparison population was matched 4:1 by sex, region of residence at birth and birth year/month. We analysed education, employment, income, job position and national insurance use (sickness benefits, parental leave benefits and social welfare provision) at 25 and 30 years of age. Results: We identified 1812 individuals with coeliac disease (6888 comparison population) at 25 years of age and 263 individuals (984 comparison population) at 30 years of age. No statistically significant differences were seen in education, employment, income, job position, use of parental leave benefits or social welfare provision. More individuals with coeliac disease used sickness benefits at age 25 years (OR 1.34 [95% CI 1.12-1.59]). Conclusion: In this register study, we showed that coeliac disease diagnosed in childhood does not cause disadvantages on career progression on a population level. However, findings suggest that coeliac disease increases the risk for sickness benefit use
Evaluation of ceftazidime/avibactam in combination with colistin against KPC-2-producing Klebsiella pneumoniae in static and dynamic time-kill experiments
No full textObjectives Ceftazidime/avibactam is used for severe infections caused by carbapenemase-producing Klebsiella pneumoniae. Combination therapy with older antibiotics is frequently used, but the supporting data are limited. This study aimed to evaluate ceftazidime/avibactam in combination with colistin against KPC-2-producing K. pneumoniae. Material and methods Five clinical KPC-2-producing K. pneumoniae strains were characterized by phenotypic antibiotic susceptibility testing and whole-genome sequencing. Single antibiotics and combinations were evaluated in 24-h static time-kill experiments with ceftazidime/avibactam concentrations of 0.5x MICratio and colistin at 0.5x and 1x MIC. One strain was subjected to 32-h dynamic time-kill experiments with ceftazidime/avibactam at concentrations mimicking patient pharmacokinetics in plasma and colistin added to 1 mg/L, i.e. the average free steady-state concentration. Population analysis was performed at 0, 16, and 32 h by plating at 4x and 8x MICratio (ceftazidime/avibactam) or MIC (colistin) to assess resistance development. Results All strains were susceptible to ceftazidime/avibactam and colistin, had mutations in ompK35 and ompK36, and carried multiple beta-lactamase genes. Ceftazidime/avibactam combined with colistin demonstrated 24-h synergy in static time-kill experiments against 3 of 5 strains. Although ceftazidime/avibactam and colistin alone showed rapid initial killing in the dynamic experiments, regrowth occurred after 4-8 h. The three-drug combination displayed a bactericidal effect and synergy at 14-32 h. Resistance development resulting in 64-fold MIC increases was observed in experiments with colistin alone. Conclusions This study showed synergy with ceftazidime/avibactam and colistin against KPC-2-producing K. pneumoniae at clinically relevant concentrations. More studies are warranted to investigate the clinical potential of this combination
Fringe Trees for Random Trees With Given Vertex Degrees
No full textWe prove asymptotic normality for the number of fringe subtrees isomorphic to any given tree in uniformly random trees with given vertex degrees. As applications, we also prove corresponding results for random labeled trees with given vertex degrees, for random simply generated trees (or conditioned Galton-Watson trees), and for additive functionals. The key tool for our work is an extension to the multivariate setting of a theorem by Gao and Wormald (2004), which provides a way to show asymptotic normality by analyzing the behavior of sufficiently high factorial moments
Observation of ¯ Production in Pb+Pb Collisions at √NN=5.02 TeV with the ATLAS Detector
No full textTop-quark pair production is observed in lead–lead (Pb+Pb) collisions at √NN=5.02 TeVat the Large Hadron Collider with the ATLAS detector. The data sample was recorded in 2015 and 2018, amounting to an integrated luminosity of 1.9 nb−1. Events with exactly one electron and one muon and at least two jets are selected. Top-quark pair production is measured with an observed (expected) significance of 5.0 (4.1) standard deviations. The measured top-quark pair production cross section is ¯=3.6 +1.0−0.9(stat) +0.8−0.5(syst) μb, with a total relative uncertainty of 31%, and is consistent with theoretical predictions using a range of different nuclear parton distribution functions. The observation of this process consolidates the evidence of the existence of all quark flavors in the preequilibrium stage of the quark-gluon plasma at very high energy densities, similar to the conditions present in the early Universe
Simulations of Neutron-Induced Light-Ion Emission using TALYS and Geant4
No full textIn this project, the limitations from the experimental setup called Medley, installed at the Neutrons For Science (NFS) facility at GANIL and used to study the production of light ions in neutron-induced reactions, are investigated. In particular, the focus of this project is to find the optimal target thickness for the targets 12C, 28Si and 56Fe that maximizes the detection of low energy emitted protons and alpha particles from neutron-induced reactions. Using Geant4 [1] we could obtain the shifted energy spectrum for the aforementioned ions, at different neutron energies and for different target thicknesses. Finally, being able to find the target’s thicknesses that maximizes the number of ions detected. We could conclude that the optimal thickness for detecting alpha particles from an Fe-56 target, at a neutron energy of 10 MeV is 25 µm. Besides that, a systematic study was done, exploring how the evaporation peak for different target elements and emitted ions evolves with neutron energies in the range of 10 to 40 MeV. For that, the reaction code TALYS[2] was used to get theoretical values for the energy spectrum of the studied light ions: protons, deuterons, tritons, He-3, and alpha particles. From this study, it is concluded that, as the neutron energies increase, the evaporation peak broadens due to the particles being emitted at several different energies and that the evaporation peak shifts to higher energy particles, meaning that higher energy particles are produced.I detta projekt undersöks begränsningarna i det experimentella systemet Medley, installerat vid Neutrons For Science (NFS) anläggningen vid GANIL, som används för att studera produktionen av lätta joner i neutroninducerade reaktioner. Projektets fokus är att hitta den optimala tjockleken för måltavlorna som består av följande material 12C-, 28Si- och 56Fe och som maximerar detektionen av lågenergetiska protoner och alfapartiklar från neutroninducerade reaktioner. Med hjälp av Geant4 [1] kunde vi få det energiskiftade spektrumet för ovan nämnda joner, vid olika neutronenergier och för olika tjocklekar på måltavlan. Slutligen kunde de tjocklekar som maximerar antalet detekterade joner identifieras. Vi kunde exempelvis dra slutsatsen att den optimala tjockleken för att detektera alfapartiklar från en 56Fe måltavla vid10 MeV neutronenergi är 25 µm. Utöver detta genomfördes en systematisk studie av hur emissionen av lågenergetiska joner sker för olika material, där neutronenergi varieras från 10 till 40 MeV. För detta användes reaktionskoden TALYS [2] för att få ut de teoretiska spektra för de studerade lätta jonerna: protoner, deuteroner, tritoner, He3 och alfapartiklar. Studien visar att spektrumet blir bredare med ökande neutronenergi, till följd av att partiklar emitteras med olika energier, samt att spektrumets toppens läge förskjuts mot högre energier, vilket innebär att partiklar med högre energi produceras
When adding a little is adding too much: : how discourse particles force immediate reanalysis and increase processing costs in under-specific contexts
No full textHighly frequent discourse particles (DPs) express speaker attitudes and guide utterance interpretation, but we still lack a satisfactory explanation of how DPs are actually processed. Some results show facilitation, while others show processing costs. Previous studies have aimed to elicit core meanings of DPs embedded in highly plausible contexts, in contrast to more unlikely contexts that force two quite different interpretations. The present study uses a novel eye-tracking experiment where DPs instead are presented in low-constraint contexts. The plausible interpretations consist of two ends of a natural scale: the state change of color that fades or becomes dirty (black to gray or white to gray). This design renders a more direct reflection of how DPs alter context interpretation. Results show that DPs induce immediate reanalysis, and this reanalysis differs in magnitude depending on the kind of DP used. We suggest that the processing of DPs involve three dimensions: i) linguistic intuition about the DP, ii) assumptions about speaker meaning and iii) contextual considerations. The results are interpreted through the communicative principle of language, under-specificity and the maxim of quantity. We also suggest that diverging results from previous studies in the field can be explained using the same analytical lens
Therapeutic options for human papillomavirus-positive tonsil and base of tongue cancer
No full textThe incidences of human papillomavirus-positive (HPV+) tonsillar and base tongue squamous cell carcinomas (TSCC and BOTSCC) have increased in recent decades. Notably, HPV+ TSCC and BOTSCC have a significantly better prognosis than their HPV-negative counterparts when treated with current surgical options, radiotherapy, or intensified chemoradiotherapy. However, a cure is not achieved in 20% of patients with HPV+ TSCC/BOTSCC. Meanwhile, cured patients often present with severe chronic side effects. This necessitates novel tailored alternatives, such as targeted therapy, immune checkpoint inhibitors (ICIs), and treatment de-escalation, together with better follow-up. Current precision medicine therefore focuses on detecting predictive and driver cancer genes to better stratify patient treatment, provide those with poor prognostic markers targeted therapy, and select those with favorable markers for de-escalated therapy. Moreover, detecting cell-free HPV DNA (cfHPV DNA) in plasma before and after treatment has been attempted to improve follow-up. In this context, this perspective discusses the significance of optimally defining HPV+ status, which requires HPV DNA and p16INKa overexpression, using prognostic markers, such as high CD8+ T-cell counts and HPV E2 mRNA expression, tumor size, and following cfHPV DNA for patient selection for specific therapies. Clinical trials with ICI with/without chemotherapy, targeted therapy with specific inhibitors—such as phosphoinositide 3-kinase and fibroblast growth factor receptor inhibitors—or immune therapy with various HPV-based vaccines for treating recurrences have yielded promising results
