Publikationsserver der Ostbayerischen Technischen Hochschule Regensburg
Not a member yet
6172 research outputs found
Sort by
Workflow for the development of a non-invasive feedback device to assess pelvic floor contractions
No full textIntroduction
Pelvic floor disorders affect about 40% of women worldwide [1]. Pelvic floor muscle (PFM) training is both a preventive and a therapeutic intervention. Current PFM training devices are invasive and have little scientific evidence. The idea is to develop a noninvasive feedback device to assess adequate PFM contraction. Therefore, evidence-based female musculoskeletal models, non-invasive data acquisition, sensor technology and artificial intelligence (AI) will be combined. This work presents the workflow to achieve such a feedback device and describes the interaction of the technologies used.
Methods
Exercises that induce PFM contractions have been evaluated and defined. Motion capture of these exercises will provide input for female musculoskeletal models. A combination of biomechanical rigid body and FEM simulations will be used to estimate PFM contractions. In addition, a non-invasive sensor will measure pelvic floor activity. The simulated and measured data will be used to develop an AI model that provides feedback on PFM contractions based on non-invasive data collection.
Results
The AMMR (AnyBody Managed Model Repository) of the AnyBody modelling system (AMS, Aalborg, Denmark) serves as the initial model for performing inverse dynamic simulations of the exercises. To calculate the PFM forces, the full-body model must be supplemented with the relevant pelvic floor structures and a mass model of the internal organs. A modified abdominal pressure model must also be incorporated. The AMS calculates the PFM activities caused by the internal organ loads and the generated abdominal pressure during the exercises. The muscle activities are transferred to a FEM model of the female pelvic floor (SfePy, simple finite elements in Python). The identical pelvic floor structures were integrated into the FEM model as in the AMS. Active PFM contractions can be simulated using the FEM model. Movement of the coccyx due to PFM contractions has been reported in the literature [2,3]. Therefore, a noninvasive coccyx motion sensor will be developed to provide additional information on PFM contractions. The measured data (coccyx motion sensor, motion capture) and the simulation results of the models will be combined to create an AI feedback model using Python. The final feedback device will consist of the AI model and the developed coccyx motion sensor, which can reproduce the resulting PFM contractions based on the sensor data and simplified motion tracking.
Discussion
The creation of the AMS and the FEM model is a prerequisite for the development of the feedback device. The relevant structures in the models are located inside the body. This limits the ability to observe the structures during the exercises, which can lead to difficulties in model validation. The development of a user-friendly sensor with sufficient measuring accuracy of the coccyx motion is another challenge. Nevertheless, the workflow represents a promising approach to develop a noninvasive feedback system to assess PFM contraction.
References
1. Wang et al, Front Public Health, 10:975829, 2022.
2. Bø et al, Neurourol Urodyn, 20:167–174, 2001.
3. Fujisaki et al, J Phys Ther Sci, 30:544–548, 2018.
Acknowledgements
This work was supported by the project no. BYCZ01-014 of the Program INTERREG Bavaria – Czechia 2021–2027
Erhebungsmaterialien im Projekt Demokratieakzeptanz und Partizipation von Geflüchteten (DePaGe)
No full textErhebungsmaterialien (Fragebogen und Informationstext) des Forschungsprojekts Demokratieakzeptanz und Partizipation von Geflüchteten (DePaGe) im Rahmen des bayerischen Forschungsverbunds zur Zukunft der Demokratie (ForDemocracy)
Kinetic cooling in mid-infrared methane photoacoustic spectroscopy: A quantitative analysis via digital twin verification
No full textThis study presents a detailed quantitative analysis of kinetic cooling in methane photoacoustic spectroscopy, leveraging the capabilities of a digital twin model. Using a quantum cascade laser tuned to 1210.01 cm⁻¹, we investigated the effects of varying nitrogen-oxygen matrix compositions on the photoacoustic signals of 15 ppmV methane. Notably, the photoacoustic signal amplitude decreased with increasing oxygen concentration, even falling below the background signal at oxygen levels higher than approximately 6 %V. This phenomenon was attributed to kinetic cooling, where thermal energy is extracted from the surrounding gas molecules rather than added, as validated by complex vector analysis using a previously published digital twin model. The model accurately reproduced complex signal patterns through simulations, providing insights into the underlying molecular mechanisms by quantifying individual collision contributions. These findings underscore the importance of digital twins in understanding the fundamentals of photoacoustic signal generation at the molecular level
Untersuchungen zum Tragverhalten von Einfach- und Mehrfachverbundankern im Zuge des BA 2/2 der A281 in Bremen
No full textIm Vorfeld zum Neubau der A281 BA2/2 in Bremen wurden u.a. 14 Eignungsanker hergestellt und das Lastabtragsverhalten in den anstehenden Wesersanden ermittelt. Die Verpresskörper wurden in Sanden mit mitteldichter und dichter Lagerung angeordnet und dabei die Länge und Anzahl der Verpresskörper systematisch variiert. Neben der Messung von Kraft und Verschiebung am Ankerkopf wurde mithilfe hochauflösender faseroptischer Instrumentierung von Verpresskörpern und Ankerlitzen der Lastabtrag über den reinen Verpresskörper hinaus über die gesamte Verbundlänge ermittelt. Anhand der Ergebnisse wird die sukzessive Aktivierung der Baugrundwiderstände sichtbar. Die Ergebnisse zeigen einen näherungsweise linearen Zusammenhang zwischen den untersuchten 6 m bis 12 m langen Verpresskörpern und den ermittelten Herausziehwiderständen im Sand und somit ein deutlich günstigeres Lastabtragsverhalten, als es bisherige Erfahrungen gezeigt haben
Online Political Participation of Refugees in Germany: Analysis of a Survey in Bavaria
No full textNew media are an important resource for the political participation of marginalized groups. However, there is still a lack of knowledge about the factors that influence whether this opportunity is used. Influencing factors emerge from both the Civic Voluntarism Model and previous research on migrants’ political activity. Using data collected in 2019/2020 from 486 refugees living in Bavaria, we estimate linear probability models to investigate facilitating factors for the use of new media for political participation. The analysis shows that refugees in Germany who inform themselves about politics online tend to be predominantly male, higher educated, and politically involved. Language skills and the duration of stay also appear to be important. For the expression of one’s political opinion online, gender and language skills have an impact, but informing oneself about German politics, political interest, and offline political activity in Germany are the most relevant factors. This indicates that there is a close link between on- und offline activity. However, using the internet requires particular resources, so that people with lower language skills, for example, are particularly disadvantaged
„Analog Discovery 2“ mobiles Praktikum OTH Regensburg
No full textAllen Studierenden wird zu Beginn des zweiten Semesters ein elektronisches Experimentierset mit einem Analog Discovery als „mobiles Praktikum“ für die gesamte Dauer des Studiums ausgeliehen. Im Rahmen der Lehrinnovationsprofessur (LIP) wird Material für möglichst viele verschiedene Module des Studiengangs entwickelt, um Fachwissen praktisch erfahrbar zu machen und gleichzeitig die Module inhaltlich zu verknüpfen
ChatGPT for futures: how large language models can support the development of future scenarios using the Cone of Plausibility
No full textRecently, large language models (LLMs), such as the “Chat Generative Pre-trained Transformer”, commonly known as ChatGPT, are substantially impacting the world of (big) data analytics and artificial intelligence. This paper explores how ChatGPT-4 can be used to support the development of future scenarios using the Cone of Plausibility method. The most recent version ChatGPT 4.0 differs from its predecessors in that it can now access real-time data to generate its responses. This makes it an attractive tool for intelligence analysts, who are faced with the permanent challenge of having to sift through large amounts of data whilst also providing actionable products in a timely manner. Previous versions of ChatGPT were prone to errors such as making up information. But what if, using the right prompts, version 4.0 could become a useful sparring partner for developing future scenarios? In this study, we will ask ChatGPT 4.0 to generate sets of plausible future scenarios for “Russia 2035+” based on previously specified key drivers and assumptions. The LLM’s potential for error is reduced because it is not working with hypotheses that it has generated itself. This could turn ChatGPT into a quick-thinking sparring partner for imagining varieties of future scenarios, supporting out-of-the box thinking and further minimising cognitive biases. To evaluate the reliability and accuracy of the results generated by ChatGPT, we will use the VV&A concept developed by the US Department of Defence. According to this concept, credibility in an intelligence product is rooted in the verification, validation and accreditation of its analysis. This process is routinely followed by every intelligence analyst. We will have found out to which extent and under which circumstances ChatGPT 4.0 can credibly support the development of future scenarios
Application Integration Framework for Large Language Models
No full textLarge Language Models (LLMs) have unlocked new opportunities in processing non-structured information. However, integrating LLM into conventional applications poses challenges due to their non-deterministic nature. This paper introduces a framework designed to effectively integrate LLM into intermediate modules by ensuring more consistent and reliable outputs. The framework includes three key components: the Sieve, which captures and retries processing of incorrect outputs; the Circuit Breaker, which stops processing persistently incorrect outputs; and the Optimizer, which enhances processing efficiency by combining inputs into single prompts. Experimental results employing structured methodology demonstrate the framework's effectiveness, achieving significant improvements a 71.05% reduction in processing time and an 82.97% reduction in token usage while maintaining high accuracy. The proposed framework, agnostic to specific LLM implementations, aids the integration of LLMs into diverse applications, enhancing automation and efficiency in fields such as finance, healthcare, and education
Considering a Unified Model of Artificial Intelligence Enhanced Social Work: A Systematic Review
No full textSocial work, as a human rights–based profession, is globally recognized as a profession committed to enhancing human well-being and helping meet the basic needs of all people, with a particular focus on those who are marginalized vulnerable, oppressed, or living in poverty. Artificial intelligence (AI), a sub-discipline of computer science, focuses on developing computers with decision-making capacity. The impacts of these two disciplines on each other and the ecosystems that social work is most concerned with have considerable unrealized potential. This systematic review aims to map the research landscape of social work AI scholarship. The authors analyzed the contents of 67 articles and used a qualitative analytic approach to code the literature, exploring how social work researchers investigate AI. We identified themes consistent with Staub-Bernasconi’s triple mandate, covering profession level, social agency (organizations), and clients. The literature has a striking gap or lack of empirical research about AI implementations or using AI strategies as a research method. We present the emergent themes (possibilities and risks) from the analysis as well as recommendations for future social work researchers. We propose an integrated model of Artificial Intelligence Enhanced Social Work (or “Artificial Social Work”), which proposes a marriage of social work practice and artificial intelligence tools. This model is based on our findings and informed by the triple mandate and the human rights framework