Publikationsserver der Ostbayerischen Technischen Hochschule Regensburg
Not a member yet
6172 research outputs found
Sort by
Examining Reliance Patterns on AI Advice in Medical Imaging: a Mixed-Methods Randomized Crossover Experiment
No full textBackground: Artificial intelligence (AI) holds significant potential to support diagnostic decision-making; however, evidence regarding its clinical utility remains mixed. Often, the collaboration between clinicians and AI systems does not surpass the individual performance of unaided humans or standalone AI. Yet, currently, the mechanisms that limit human-AI synergy are poorly understood. This study examined the impact of AI advice on diagnostic decision-making among experts and novices, focusing on reliance patterns.
Methods: We used a mixed-methods crossover experimental design with a think-aloud and an eye-tracking study arm. Participants were 50 task experts (radiologists) and 75 novices (non-radiologist physicians and medical trainees) from 10 countries. They reviewed 50 head CT scans and every case was examined in three time-separate sessions in randomized order. In each session, participants were exposed to different experimental conditions: (a) control, no AI prediction; (b) basic advice, AI prediction without annotations; and (c) XAI advice, AI prediction with scan annotations. For each case, participants had to determine if the patients had an intracranial hemorrhage (ICH). The main outcomes were diagnostic performance, confidence in the diagnosis, case reading time, and AI advice usefulness ratings.
Findings: Both overreliance on incorrect advice and underreliance on correct advice occurred. Underreliance was associated with high uncertainty and, in absolute terms, had a more detrimental impact on diagnostic performance than overreliance. Correct XAI advice reduced underreliance, improved performance (OR=1·84, p<0·0001), and confidence (b=0·15, p<0·0001), particularly when reviewing more difficult cases with ICH. Surprisingly, correct XAI did not reduce reading time (b=1·81, p=0·0713). XAI was perceived as more useful than basic AI advice (b=0·12, p=0·0029), especially among novices.
Interpretation: The occurrence of both under- and overreliance highlights the need for efficient counterstrategies beyond classic XAI methods to foster appropriate reliance and synergy
Untersuchungen und Analyse des Verhaltens von Böden und Boden-Bindemittel-Gemischen unter zyklischer Belastung
No full textWährend es zahlreiche Untersuchungen zum Verhalten von Böden unter zyklischen und dynamischen Belastungen gibt, gibt es nur wenige Veröffentlichungen zu zyklischen Belastungen an Bindemittel-stabilisierter Böden. Dabei werden diese Boden-Bindemittel-Gemische (BBG) vielseitig eingesetzt (z.B. qualifizierte Bodenverfestigung, Bodenmischsäulen, zeitweise fließfähige Erdbaustoffe).
Die durchgeführten Versuche zielen darauf ab, ein tiefergehendes Verständnis für das Materialverhalten von BBG zu gewinnen. Hierfür wurden an verschiedenen Mischungen zyklische Versuche mit variierenden Spannungsamplituden durchgeführt.
Die bisher durchgeführten Versuche untersuchen sowohl Zyklen im niedrigen Lastbereich als auch im hohen Lastbereich nahe der einaxialen Druckfestigkeit.
Gängige Modelle zur Beschreibung des Setzungsverhaltens in Folge von zyklischer Einwirkung werden vorgestellt und mit den Ergebnissen für BBG verglichen
Challenges and Approaches for Long-Term Reactive Power Forecasting in Power Systems
No full textThe transition from fossil fuels to renewable energy sources is driving profound changes across all voltage levels of the electrical grid, altering active power (P) as well as reactive power (Q) behavior. This transformation involves a shift from predominantly top-down, unidirectional electricity flows to bidirectional and increasingly complex flow patterns, fundamentally altering the operation and utilization of power systems. Many countries are already reporting notable changes in their transmission grid Q. This can lead to increased costs for grid expansion and compensation equipment, as well as grid losses. To minimize these losses and ensure grid stability and efficiency, understanding and forecasting Q long-term becomes essential. This paper addresses the challenges of long-term (1-20 years) Q forecasting in general by analyzing Q behavior in grids, the main influencing grid parameters, and reviewing existing forecasting approaches. It evaluates the applicability of these approaches and highlights current research gaps and trends. By focusing on this underexplored area, the paper aims to support the development of more advanced and reliable long-term Q forecasting techniques
Michael Weber: Werkstätten für behinderte Menschen - Inklusionshemmnis oder Weg zur Teilhabe?
No full textChristina Hansen, Kathrin Plank (Hrsg.): Im Zwiegespräch mit Janusz Korczak. Was wir im Europa des 21. Jahrhunderts vom „Weltenwanderer, Tabubrecher und Kräftezähmer“ lernen können
No full textLernen in digitalen Realitäten : XR als Chance für die Pflegebildung
No full textExtended Reality (XR) eröffnet neue Möglichkeiten,den Theorie-Praxis-Transfer in der Pflege zu ver-bessern. Der Beitrag zeigt, wie Virtual Reality (VR),Augmented Reality (AR) und Mixed Reality (MR)Lernprozesse unterstützen, welche pädagogischenKompetenzen Lehrende benötigen und welche He-rausforderungen bei der Implementierung zu be-achten sind
Abstract: DIY Challenge Blueprint
No full textThe high cost of challenge platforms prevents many people from organizing their own competitions. The do-it-yourself (DIY) challenge blueprint [1] allows you to host your own biomedical AI benchmark challenge. Our DIY approach circumvents the current constraints of commercial challenge platforms. A sovereign, extensible and cost-efficient deployment is provided via containerised, identity-managed and reproducible pipelines. Focus lies on GDPR-compliant hosting via infrastructure-as-code, automated evaluation, modular orchestration, and role-based identity and access management. The framework integrates Docker-based execution and standardised interfaces for task definitions, dataset curation and evaluation. All in all it is designed to be flexible and modular, as demonstrated in the MICCAI 2024 PhaKIR challenge [2, 3]. In this case study, different medical tasks on a multicentre laparoscopic dataset with framewise labels for phases and spatial annotations for instruments across fulllength videos were supported. This case study empirically validates the DIY challenge blueprint as a reproducible and customizable challenge-hosting infrastructure. The full code can be found at https://github.com/remic-othr/PhaKIR_DIY
Optimal Reactive Power Planning by covering deficient voltage demands with additional reactive power sources using a sensitivity approach
No full textThe determination of the optimal location, types and size of additional required reactive power sources is a main interest of Reactive Power Planning (RPP) investigations and of this paper. The planning problem addresses the maintenance of voltage stability, which is an ancillary service in Germany, and can be formulated as a mixed-integer optimization prob-lem. The aim of this work is a techno-economically efficient coverage of the deficient steady state voltage deviation demands (VDD) by placing new reactive power sources with adequate size or extending the reactive power potential of existing sources. An optimization model is set up to solve the planning problem by using load flow voltage sensitivities, which quantify and evaluate the technical efficiency of new reactive power sources in relation to the present voltage deviation demands. Operating and investment costs are used as economic decision parameters. Within an application example, which is carried out with an exemplary transmission system, the solution of the planning problem is discussed
Methodology for long-term reactive power forecasts of medium-voltage grids
No full textUnderstanding and assessing the long-term behavior of reactive power is crucial for power system planning and invest-ment decisions. This is particularly growing relevant for the lower voltage-grid levels, which are gaining importance due to significant changes in grid behavior, operational strategies, energy generation and consumption patterns. Currently, standardized methodologies for forecasting reactive power in medium-voltage grids are lacking, creating a gap in research and practical applications. To optimize grid investments and expansions, informed decisions based on reliable research and accurate forecasts are essential. This paper presents a methodology for long-term reactive power forecasting in me-dium-voltage grids, introducing a new approach to account for low- and medium-voltage grid interactions while address-ing common challenges like limited available grid data and scenario framework definition. The aim of the forecast is the future reactive power behaviour of a medium-voltage-grid at the high-voltage side of the high-voltage to medium-voltage transformer. The medium-voltage-level is modelled using a grid model approach, while low-voltage grids are being mod-elled using generic distribution grids. A 209-bus medium-voltage-grid model is used as an application example and to provide two exemplary use cases for a reactive power forecast for 2045
