Hybrid Open Access - OA-Veröffentlichung in einer Subskriptionszeitschrift/-medium
Refine
Year of publication
Document Type
- Article (193)
- conference proceeding (article) (2)
- Part of a Book (1)
- conference talk (1)
- conference proceeding (volume) (1)
- Review (1)
Is part of the Bibliography
- no (199)
Keywords
- Photoacoustic spectroscopy (5)
- Shotcrete (4)
- 3D volumetry (3)
- Artificial Intelligence (3)
- Biomechanik (3)
- Durability (3)
- Maschinelles Lernen (3)
- Medical Image Computing (3)
- Acetone breath analysis (2)
- Acetone detection (2)
Institute
- Fakultät Maschinenbau (62)
- Fakultät Informatik und Mathematik (59)
- Fakultät Elektro- und Informationstechnik (29)
- Regensburg Center of Biomedical Engineering - RCBE (29)
- Regensburg Center of Health Sciences and Technology - RCHST (23)
- Fakultät Sozial- und Gesundheitswissenschaften (20)
- Fakultät Bauingenieurwesen (17)
- Labor Regensburg Medical Image Computing (ReMIC) (17)
- Fakultät Angewandte Natur- und Kulturwissenschaften (15)
- Labor Biomechanik (LBM) (15)
Begutachtungsstatus
- peer-reviewed (189)
Purpose
Recognizing previously unseen classes with neural networks is a significant challenge due to their limited generalization capabilities. This issue is particularly critical in safety-critical domains such as medical applications, where accurate classification is essential for reliability and patient safety. Zero-shot learning methods address this challenge by utilizing additional semantic data, with their performance relying heavily on the quality of the generated embeddings.
Methods
This work investigates the use of full descriptive sentences, generated by a Sentence-BERT model, as class representations, compared to simpler category-based word embeddings derived from a BERT model. Additionally, the impact of z-score normalization as a post-processing step on these embeddings is explored. The proposed approach is evaluated on a multi-label generalized zero-shot learning task, focusing on the recognition of surgical instruments in endoscopic images from minimally invasive cholecystectomies.
Results
The results demonstrate that combining sentence embeddings and z-score normalization significantly improves model performance. For unseen classes, the AUROC improves from 43.9% to 64.9%, and the multi-label accuracy from 26.1% to 79.5%. Overall performance measured across both seen and unseen classes improves from 49.3% to 64.9% in AUROC and from 37.3% to 65.1% in multi-label accuracy, highlighting the effectiveness of our approach.
Conclusion
These findings demonstrate that sentence embeddings and z-score normalization can substantially enhance the generalization performance of zero-shot learning models. However, as the study is based on a single dataset, future work should validate the method across diverse datasets and application domains to establish its robustness and broader applicability.
Objectives
The study evaluated the influence of digital design software on crown morphology, occlusal characteristics, fracture force, and marginal fit across varying preparation designs for an identical target tooth.
Methods
A resin-based tooth (tooth 36) was digitized, manufactured (
), individually prepared and re-digitized. Five design groups were established using conventional software proposals, technician designs, two AI-based software solutions, and natural tooth-based reference designs. All systems employed consistent parameters. Crown designs were digitally assessed using quantitative morphological and occlusal metrics in reference to the original tooth. Crowns were milled, marginal fit was measured via digital microscopy, and fracture resistance was determined after thermal cycling and mechanical loading.
Results
Morphological metrics revealed statistically significant deviations across groups, with the technician design achieving the best performance. Occlusal metrics showed high deviations in the positional accuracy of the contact points across all groups. Technician and AI-based designs exhibited comparable functional results. None of the design groups were able to achieve contact with all relevant antagonist teeth, due to high deviations in the mesiolingual cusp. Conventional software designs exhibited the lowest fracture forces. Significant improvements were achieved through technician intervention. Vertical marginal discrepancies remained comparable across groups.
Significance
Improved functional and morphological design combined with high fracture resistance can reduce the need for clinical adjustments, minimize wear, and enhance crown longevity. Digital design software significantly influences crown morphology, occlusal characteristics and fracture forces. Vertical marginal discrepancies remain similar. AI-driven approaches demonstrate comparability with technician designs in terms of fracture forces, functional performance, and marginal fit.
PV and wind systems with PEM electrolysis offer great potential for producing hydrogen with low emissions. Our research has identified the ecologically optimal size of PEM in relation to fixed PV/wind capacities. We calculate efficiencies and production volumes for PEM with 240 capacity and site variations. We analyse the global warming potential of all systems and draw conclusions about the optimal system design. The lowest GWP is achieved at the site with the highest full load hours with 1.32 kg CO2-eq/kg H2 (Wind, 28 MW electrolysis) and 4.24 kg CO2-eq/kg H2 (PV, 23 MW electrolysis). We have identified a clear trend: increasing PV/wind full load hours leads to higher ideal PEM capacities. However, there is a significant discrepancy between the ideal economic and ecological capacity. Furthermore, higher electrolysis capacities can achieve lower emissions as they increasingly operate at a more efficient partial load.
In recent years, More-Than-Human-Design (MTHD) has gained traction in HCI, as exemplified by a growing body of workshops and publications. As this nascent field progresses, unresolved conceptual and practical methodological challenges continue to surface. It is time for MTHD researchers to critically revisit underlying assumptions, concepts and processes that shape their work. This one-day workshop invites the DIS community to engage in two key areas: a critical discussion on anthropocentrism, a core yet multifaceted concept in MTHD, and the presentation of practical work that either reveals paradoxes within MTHD – concerning how to decenter humans and how to engage with more-than-human actors – or explores constructive ways to navigate these paradoxes. Through mapping and discussing the challenges, we aim to identify patterns and shared concerns. The workshop will culminate in an open-ended Archive of Tensions, accessible via the workshop website, serving as a collective resource for ongoing reflection and inquiry. Furthermore, the Archive will be presented in an online event and the results will be published as an article.
In this study, the Mg alloy WE43 was solution annealed and precipitation hardened prior to electropolishing to evaluate the effects of different microstructures on the electropolishing result. While coarsely distributed precipitates led to surfaces showing wavy structures and dents after electropolishing, a uniform microstructure resulted in an even finish of the surface. The homogenization and refinement of the microstructure by heat treatment is a method to ensure improved electropolished surfaces of Mg materials
Road freight transportation accounts for a great share of the anthropogenic greenhouse gas (GHG) emissions. In order to provide a common methodology for carbon accounting related to transport activities, the European Committee for Standardization has published the European Norm EN-16258. Unfortunately, EN-16258 contains gaps and ambiguities and leaves room for interpretation, which makes the comparison of the environmental performance of different logistics networks still difficult and hinders the identification of best practices. This research contributes to the identification of particularly meaningful principles for the allocation of GHG to shipments in road freight transportation by presenting an analytical framework for studying the performance of the EN-16258 allocation schemes with respect to accuracy, fairness, and the GHG minimizing incentive. In doing so, we continue previous studies that analyzed two important aspects of the EN-16258 allocation rules: accuracy and fairness. This study provides further insights into this allocation problem by investigating the incentive power of the different allocation schemes to opt for the GHG minimal way of running a road freight network. First, we complement the list of transport scenarios introduced in prior studies and present two novel scenarios. Second, we carry out a series of numerical experiments to compare the EN-16258 allocation rules with respect to accuracy, fairness, and the GHG minimizing incentive. We find that the results may differ significantly for the two scenarios, suggesting a case-by-case recommendation. This is particularly interesting because the first scenario confirms the results of the prior studies, while the second scenario rather contradicts them.
An inverse optimization approach for studying sustainability preferences in sourcing decisions
(2024)
Throughout many societies around the globe, there is growing awareness of the urgent need for the transition towards a sustainable economy. Research shows that buying firms have substantial leverage to initiate sustainable development by controlling the sustainable performance of their suppliers. In that context, this article presents a novel methodology based on inverse optimization to derive the implicit preferences of decision-makers in the trade-off between traditional sourcing objectives and sustainability in the supplier selection and order allocation process. The derived implicit preferences can then be used for further analyses to gain a better understanding of the characteristics of purchasing managers and sourcing situations that come with particularly high/low preferences placed on sustainability. Since the inverse optimization approach is computationally resource-intensive and consumes a significant amount of time, we present a scalable state-of-the-art cloud architecture that allows solving an arbitrary number of optimization programs in an acceptable amount of time. We demonstrate the feasibility of the proposed methodology in a real-world case. In doing so, we test how important sustainability aspects are in the supplier selection and order allocation decisions of one of the world's largest automotive parts manufacturers.
AIM:
To develop a 3D-printed model enabling treatment simulation of trauma-related endodontic and restorative procedures like revitalization or placement of an apical plug on teeth with incomplete root formation.
METHODOLOGY:
To generate a realistic training model suitable for trauma-related simulations, CT-Data sets were segmented, combined, adjusted and optimized using computer-aided design features. Specific focus was on optimally reflecting characteristics of immature teeth like open apical foramina and thin root walls. For reflection of the revitalization process, a set-up including a tube filled with red liquid under pressure was developed to be perforated during the procedure by the trainee, simulating bleeding from the apical papilla into the root canal system. Design was based on a combination of cost-effective simple parts combined with 3D-printed components to achieve maximal accessibility, exchangeability and re-usability. Model assembly and preparation were described in step-by-step instructions.
RESULTS:
As a result of the developmental process, the presented model qualifies for the training of the endodontic management of complications on immature teeth after dental trauma. Step-by-step descriptions aligned with the clinical procedures are performable for a revitalization procedure, for pulpotomy followed by fragment re-attachment and splint application and for placement of an apical plug using hydraulic calcium silicate cements.
CONCLUSIONS:
The model successfully achieved its intended objectives, so that it may be used in the future in various dental trauma treatment simulations for undergraduate and postgraduate education on the endodontic management of trauma-related therapies in immature teeth. Its common manufacturing process, affordability and re-usability support accessibility and sustainability.
Photodynamic therapy (PDT) is a minimally invasive cancer treatment method. The antitumor effect of PDT mediated by the photosensitizer meta-tetra(hydroxyphenyl)chlorin (m-THPC) was studied in vitro on HER2-positive (HER2+) breast cancer cells. Since the application of m-THPC is limited due to poor tumor targeting abilities, the polymer carrier dextran-grafted-polyacrylamide (D-g-PAA) with attached gold nanoparticles was loaded with m-THPC to provide a light-responsive, tumor targeting nanocomposite. The nanocomposite was absorbed by the cancer cells within 90 min and activated through irradiation with red light at a wavelength of 660 nm. In this study, the efficiency and ideal dose of the nanocomposite were evaluated and determined based on cell viability and induction of apoptosis of HER2+ breast cancer cells and human breast epithelial cells. We demonstrate that the nanocomposite exhibited phototoxicity with cell mortality exceeding 50% for the HER2+ breast cancer cells. In contrast, the human breast epithelial cells were minimally affected at the same concentration of the nanocomposite, demonstrating the tumor-targeting abilities of the developed nanocomposite. Dark toxicity was also comparably lower, indicating the limited effect of the unactivated nanocomposite. The results revealed that apoptosis was the primary mode of cell death in HER2+ breast cancer cells. Apoptotic cell death was induced in HER2+ breast cancer cells via the caspase-3, caspase-9, and Bax pathways. In conclusion, the studied nanocomposite demonstrated excellent PDT performance with good antitumor activity under light irradiation in vitro and is worthy of further investigation as an alternative treatment for HER2+ breast cancer.
Bifacial solar photovoltaic (PV) technology is currently taking over the solar PV module market, exceeding a 90% share in 2025. This important technology must be included in energy system modelling. This study provides a method for calculating the yield of monofacial and bifacial power plants in fixed-tilted, single-axis tracking, and east-west facing vertical setup. A novel method is introduced to maximise the capacity density of solar PV power plants without the need for detailed land cost for the most efficient use of the occupied area. The results indicate a 15-20% yield gain from single-axis tracking compared to fixed-tilted power plants, and a limited bifacial gain of up to 10% for most areas of the world. Higher bifacial gains are sporadically possible in specific conditions. Fixed-tilted systems show higher bifacial gains. Optimising tilt angles and row pitch would allow for 147 MW/ km2 capacity density today, though on average 70-110 MW/km2 can be achieved for 20.2% module efficiency. The impact on the power system, studied in a free cost optimisation scenario and forcing vertical bifacial PV scenario, implying agrivoltaics, is not significant with a +/- 10% change in total solar PV capacity, change in installed wind power of on average-10%, increase of installed battery capacity of on average 5%, and an on average changed levelised cost of electricity of-2% globally. Bifacial solar PV technology has been found to be beneficial but no game changer for future power systems; system improvements are widely possible underlining the important role of this technology.