Publikationsserver der Ostbayerischen Technischen Hochschule Regensburg
Not a member yet
6172 research outputs found
Sort by
Underreliance Harms Human-AI Collaboration More Than Overreliance in Medical Imaging
No full textImportance: The use of artificial intelligence (AI) to support clinicians in diagnostic decision-making holds significant potential; however, evidence regarding its clinical utility remains mixed. In many cases, the interaction between healthcare professionals and AI systems does not improve collaborative performance compared to the standalone performance of humans or AI. Currently, the underlying mechanisms that limit human-AI collaboration are poorly understood. Objective: To examine the impact of AI advice on diagnostic decision-making among experts and novices, focusing on understanding the role of explainability (XAI) on users’ reliance on advice. Design, Setting, and Participants: A mixed-methods design combining a crossover experimental design with a think-aloud and an eye-tracking study arm was conducted in 2023. Participants were task experts (radiologists) and novices (non-radiologist physicians and medical trainees) from 10 countries, with the think-aloud and eye-tracking conducted in Germany. Intervention: Participants reviewed 50 patient cases containing head CT scans and patient information. Every case was reviewed in three time-separate sessions in randomized order. In each session, participants were exposed to a different experimental condition: (a) control, i.e., no AI prediction presented; (b) basic advice, i.e., AI prediction without annotations; and (c) XAI advice, i.e., AI prediction with annotations. For each case, participants had to determine if the patients had an intracranial hemorrhage (ICH), rate their confidence, and, if applicable, the usefulness of the AI advice. Main Outcome(s) and Measure(s): Diagnostic performance, confidence in the diagnosis, case reading time, and AI advice usefulness ratings. Results: The data analysis included 125 participants. The mean age was 28.5 years (SD = 6.72), and 55.2% identified as female. Underreliance on correct AI advice was associated with high uncertainty and had a more detrimental impact on diagnostic performance than overreliance on incorrect advice. XAI advice reduced underreliance and improved performance and confidence, particularly when reviewing more difficult cases with ICH. AI advice, particularly XAI, did not reduce reading time. XAI was perceived as more useful than basic AI advice, especially among novices. Conclusions and Relevance: Our findings indicate that underreliance on AI might be more harmful than overreliance, highlighting the need to develop efficient counterstrategies beyond current XAI methods
Evaluation of the creep strength of samples produced by fused deposition modeling
No full textData sheets for 3D printing materials typically include softening temperature, impact strength, tensile strength, and stiffness. However, creep strength, an important parameter for components used over an extended period, is usually not included. Nevertheless, this parameter is of significant importance for components that are used over an extended period of time.This study compares the long-term creep behavior of a selection of materials that are commonly used in fused deposition modeling 3D printing. The materials under investigation are acrylonitrile butadiene styrene, acrylonitrile styrene acrylate, polylactic acid, and polycarbonate. In addition, the influence of fiber reinforcements on these materials is also examined. A simple, reproducible test procedure is proposed for users to determine and compare creep resistance of materials. This enables developers to select materials suitable for their own requirements on creep resistance and allows 3D-printing users to compare different materials. Results suggest that fiber reinforcement generally improves creep stability in 3D-printing materials, with GreenTEC Pro Carbon and add:north PC Blend HT LCF showing the most promise in this study
Insert-Only versus Insert-Delete in Dynamic Query Evaluation
No full textWe study the dynamic query evaluation problem: Given a full conjunctive query and a sequence of updates
to the input database, we construct a data structure that supports constant-delay enumeration of the tuples in
the query output after each update.We show that a sequence of insert-only updates to an initially empty database can be executed in total time O( w( ) ), where w() is the fractional hypertree width of . This matches the complexity of the
static query evaluation problem for and a database of size . One corollary is that the amortized time per
single-tuple insert is constant for -acyclic full conjunctive queries. In contrast, we show that a sequence of inserts and deletes can be executed in total time eO( w( b ) ), where b is obtained from by extending every relational atom with extra variables that represent the “lifespans” of tuples in the database. We show that this reduction is optimal in the sense that the static evaluation runtime of b provides a lower bound on the total update time for the output of . Our approach achieves amortized optimal update times for the hierarchical and Loomis-Whitney join queries
Explainable AI for Mixed Data Clustering
No full textClustering, an unsupervised machine learning approach, aims to find groups of similar instances. Mixed data clustering is of particular interest since real-life data often consists of diverse data types. The unsupervised nature of clustering emphasizes the need to understand the criteria for defining and distinguishing clusters. Current explainable AI (XAI) methods for clustering focus on intrinsically explainable clustering techniques, surrogate model-based explanations utilizing established XAI frameworks, and explanations generated from inter-instance distances. However, there exists a research gap in developing post-hoc methods that directly explain clusterings without resorting to surrogate models or requiring prior knowledge about the clustering algorithm. Addressing this gap, our work introduces a model-agnostic, entropy-based Feature Importance Score for continuous and discrete data, offering direct and comprehensible explanations by highlighting key features, deriving rules, and identifying cluster prototypes. The comparison with existing XAI frameworks like SHAP and ClAMP shows that we achieve similar fidelity and simplicity, proving that mixed data clusterings can be effectively explained solely from the distributions of the features and assigned clusters, making complex clusterings comprehensible to humans
Optimal distributed generation placement in radial distribution system using particle swarm optimization
No full textThis paper addresses losses of power in Radial Distribution Systems (RDS), which significantly affect voltage levels and operational costs. The main aim is to optimize the positioning and size of Distributed Generation (DG) units, including Photovoltaic (PV) cells, and Wind Turbines (WT), to analyze the effect of DG placement in reducing the losses in power and enhancing the voltage profiles in radial distribution system (IEEE 33 bus system). This research employs Particle Swarm Optimization (PSO), a robust algorithm well-suited for tackling non-linear optimization issues in order to identify the appropriate placement and size for DG units. A number of scenarios with varying numbers of DG units are simulated, indicating significant reductions in active as well as reactive power losses. Likewise, the consistency and reliability requirements of modern distribution systems are improved as the voltage profile is improved. The key findings demonstrate that optimal DG integration enhances system efficiency, contributes to operational cost savings, and improves grid stability. PSO was chosen for its ability to effectively balance computational effort while achieving high accuracy in minimizing the losses associated with power and enhancing voltage profiles. In contrast to traditional optimization techniques, PSO offers superior accuracy and efficiency in addressing the challenges of non-linear optimization in RDS
Die Auswirkungen der mechanischen Kräfte eines innovativen Operationsverfahrens für die intraoperative Gewebezubereitung bei autologem Fetttransfer auf das Fettgewebe und dessen Stammzellen
No full textPrediction intervals for future Pareto record claims
No full textStochastic models and methods for quantifying extreme events are of interest in numerous disciplines. Within this paper, statistical prediction of extreme claims or losses in insurance industry is considered based on upper record values, which describe successively largest observations in a sequence of data over time. The problem of predicting a future record value (here in particular, a future record claim) based on a sequence of previously observed record values (here, past record claims) is addressed by means of prediction intervals. For an underlying Pareto distribution, respective exact and approximate intervals from the literature are summarized and modified and new ones are developed. In a simulation study, these prediction intervals are evaluated and compared regarding coverage frequency and length. The impact of the number of observed record values as well as the choice of the Pareto distribution is discussed. In the case of a small number of record values, the use of k-th record values is considered as an option for statistical analyses to predict, e.g., second largest record claims. Selected prediction methods are applied to several real data sets, which turn out to perform well and to be able to capture the magnitude of future record claims, even for fairly small numbers of record observations. For comparison, generalized Pareto distributions are fitted to real data sets and a corresponding point predictor as well as a respective upper prediction interval for the next record to appear are derived and evaluated
Herzschrittmacher, Organ oder Gehstock? Vorher-Nachher-Befragung zur subjektiven Wahrnehmung eines Gerätes zur Tiefen Hirnstimulation
No full textAIM OF THE STUDY
This study was conducted in a pre-post design with a survey of patients who had undergone deep brain stimulation (DBS) as treatment for a neurological movement disorder. The aim of the study was to compare patients' expectations and beliefs before a DBS intervention with patients' subjective experience of this intervention.
METHODOLOGY
The longitudinal study of patients (n=132) with an indication for DBS therapy was based on a written survey at the time points of preoperative screening (pre-op) and one-year follow-up (post-op).
RESULTS
Preoperatively, a clear majority of respondents believed DSB to be similar to a pacemaker intervention, but one year after the intervention less than one third did so, as they compared DBS to using a walking stick or glasses.
CONCLUSION
The experience of DBS in the patient's own body seems to be comparable by means of individually different associations, whereby the comparison with non-invasive aids predominates postoperatively. The discussion of these descriptions in the educational interview can contribute to a realistic horizon of patients' expectations before DBS.
ZIEL DER STUDIE
Es wird eine Studie im Prä-Post-Design mit einer Befragung von Patient*innen vorgestellt, bei denen zur Therapie einer neurologischen Bewegungsstörung die Indikation für eine Tiefe Hirnstimulation (THS) gegeben war. Die Patient*innen wurden vor und nach der THS-Operation zu ihren präoperativen Assoziationen bzw. postoperativen Wahrnehmungen im Umgang mit einem Hirnschrittmacher befragt. Ziel der Untersuchung ist, die bisherige Praxis der ärztlichen Aufklärung vor einer THS-Intervention um Angaben zur subjektiven Erlebnisqualität dieser Intervention ergänzen zu können.
METHODIK
Es wurde eine Längsschnittstudie mit Patient*innen (n=132) durchgeführt, in der die Teilnehmenden zweizeitig schriftlich zu Imaginationen zu ihrer subjektiven Erlebnisqualität von THS befragt wurden. Die Untersuchungen erfolgten beim präoperativen Screening (Prä-OP) sowie beim Ein-Jahres-Follow-Up (Post-OP).
ERGEBNISSE
Präoperativ gab die Mehrheit der Befragten die Assoziation Herzschrittmacher an. Postoperativ wählten nur weniger als ein Drittel diese Option, gleich häufig wie die Assoziationen Gehstock oder Brille.
SCHLUSSFOLGERUNG
Die Erlebnisqualität von THS im eigenen Körper scheint sich mittels individuell verschiedener Assoziationen vergleichen zu lassen, wobei postoperativ der Vergleich mit nicht invasiven Hilfsmitteln überwiegt. Im Aufklärungsgespräch kann die Thematisierung dieser Beschreibungen bei den Patient*innen zu einem realitätsnahen Erwartungshorizont vor THS beitragen
Gradual error detection technique for non-destructive assessment of density and tensile strength in fused filament fabrication processes
No full textFused filament fabrication (FFF) is a widely used additive manufacturing process for producing functional components and prototypes. The FFF process involves depositing melted material layer-by-layer to build up 3D physical parts. The quality of the final product depends on several factors, including the component density and tensile strength, which are typically determined through destructive testing methods. X-ray microtomography (XCT) can be used to investigate the pore sizes and distribution. These approaches are time-consuming, costly, and wasteful, making it unsuitable for high-volume manufacturing. In this paper, a new method for non-destructive determination of component density and estimation of the tensile strength in FFF processes is proposed. This method involves the use of gradual error detection by sensors and convolutional neural networks. To validate this approach, a series of experiments has been conducted. Component density and tensile strength of the printed specimens with varying extrusion factor were measured using traditional destructive testing methods and XCT. The cumulative error detection method was used to predict the same properties without destroying the specimens. The predicted values were then compared with the measured values, and it was observed that the method accurately predicted the component density and tensile strength of the tested parts. This approach has several advantages over traditional destructive testing methods. The method is faster, cheaper, and more environmentally friendly since it does not require the destruction of the product. Moreover, it facilitates the testing of each individual part instead of assuming the same properties for components from one series. Additionally, it can provide real-time feedback on the quality of the product during the manufacturing process, allowing for adjustments to be made as needed. The advancement of this approach points toward a future trend in non-destructive testing methodologies, potentially revolutionizing quality assurance processes not only for consumer goods but various industries such as electronics or automotive industry. Moreover, its broader applications extend beyond FFF to encompass other additive manufacturing techniques such as selective laser sintering (SLS), or electron beam melting (EBM). A comparison between the old destructive testing methods and this innovative non-destructive approach underscores the possible fundamental change toward more efficient and sustainable manufacturing practices. This approach has the potential to significantly reduce the time and cost associated with traditional destructive testing methods while ensuring the quality of FFF-manufactured products
NoiSLU: a Noisy speech corpus for Spoken Language Understanding in the Public Transport Domain
No full textThe use of local public transport requires the barrier-free purchase of a ticket. Travellers who are not proficient in the local language benefit from a multilingual human(ticket)machine voice interaction. This paper presents a nearly parallel audio dataset with 13218 annotated user queries from 20 speakers for English, German and Dutch. The domain-specific speech corpus can be understood as an evaluation dataset for future research in Spoken Language Understanding (SLU) and thus, it enables researches to improve the quality of human-machine interaction applications. Furthermore, we compare the SLU performance of different compositions of Automatic Speech Recognition (ASR) and Natural Language Understanding (NLU) models in baseline experiments on different test datasets