11163 research outputs found
Sort by
Stratification of the re-identification risk of medical data
No full textBedingt durch die fortschreitende Digitalisierung und technische Weiterentwicklung fallen im Rahmen des klinischen Alltags bzw. im Gesundheitswesen allgemein medizinische Daten in immer größerem Umfang an (Rüping et al. 2019). In diesen Datenbeständen werden mannigfaltige Potenziale gesehen und die Möglichkeit, mittels einer systematischen Analyse und Aufbereitung Erkenntnisgewinne für die weitere Forschung und Versorgung generieren zu können, erweckt große Erwartungen (Summa et al. 2020).
Bei aller Euphorie und Begeisterung über den möglichen Nutzen, der aus medizinischen Daten generiert werden könnte, darf jedoch keinesfalls unbeachtet bleiben, dass mit der orchestrierten Verarbeitung und Bereitstellung von Daten Datenschutzrisiken verbunden sind. Insbesondere die Evaluation des Re-Identifikationsrisikos, also der Möglichkeit der Rekonstruktion eines Personenbezugs eines anonymisierten Datensatzes, ist als kritisch anzusehen. Derzeit ist keine standardisierte Methode zur umfassenden Quantifizierung dieses Risikos für medizinische Daten bekannt.
Ziel dieser Arbeit ist die Entwicklung eines heuristischen Ansatzes zur Stratifizierung des Re-Identifikationsrisikos medizinischer Datensätze. Dazu wird ein integrativer Ansatz verfolgt, der das Risiko der Re-Identifikation nicht ausschließlich monokausal betrachtet, sondern als komplexes System unterschiedlicher Komponenten modelliert. Eingangs werden dazu mit Hilfe einer systematischen Literaturanalyse Faktoren ermittelt und beschrieben, die im Kontext von Re-Identifikationsangriffen und entsprechenden Risikoabschätzungen eine tragende Rolle spielen. Im Anschluss erfolgt eine mehrschrittige Modellbildung; die festgestellten Faktoren werden dabei zunächst abstrahiert und zu generalisierten Faktorengruppen zusammengefasst, die im weiteren Verlauf als Perspektiven bezeichnet werden. Für jeden Faktor werden in diesem Zusammenhang Kriterien vorgestellt, welche eine Einordnung der Ausprägung der jeweiligen Komponente ermöglichen. Anhand von Entity-Relationship-Modellen werden dann die Wechselwirkungen veranschaulicht, welche zwischen den einzelnen Faktoren existieren. Im letzten Schritt werden die ermittelten Wechselwirkungen zur Konstruktion einer sog. Einflussmatrix genutzt, die eine quantitative Beschreibung der Richtung und Ausprägung der zwischen den einzelnen Faktoren bestehenden Beziehungen erlaubt. Basierend auf dieser Matrix werden verschiedene Indizes vorgestellt, die zur Einordnung und Stratifizierung des Re-Identifikationsrisikos einer zu begutachtenden Datensammlung und als Grundlage für zielgerichtete Sicherheits-maßnahmen genutzt werden können.
Zum Abschluss werden zwei Gedankenexperimente beschrieben, deren Ziel in der Beurteilung der Güte der Heuristik besteht. Anhand publizierter, verifizierter Re-Identifikationsangriffe wird dabei die Plausibilität und Trennschärfe der Risiko-stratifizierung bestätigt; es zeigt sich in diesem Zusammenhang allerdings eine eher etwas geringere Einschätzung des Risikos, als es vermutlich tatsächlich war. Weitere Forschungsanstrengungen sind daher notwendig.As a result of advancing digitalization and technical development, an ever-increasing amount of medical data is being generated in everyday clinical practice and in the healthcare sector in general (Rüping et al. 2019). A wide range of potential is seen in these data sets and the possibility of generating knowledge gains for further research and care through systematic analysis and processing raises great expectations (Summa et al. 2020).
However, despite all the euphoria and enthusiasm about the potential benefits that could be generated from medical data, the data protection risks associated with the orchestrated processing and provision of data must not be ignored. In particular, the evaluation of the re-identification risk, i.e. the possibility of reconstructing a personal reference to an anonymized data set, must be regarded as critical. There is currently no known standardized method for comprehensively quantifying this risk for medical data.
The aim of this work is to develop a heuristic approach to stratify the re-identification risk of medical datasets. For this purpose, an integrative approach is pursued that does not consider the risk of re-identification exclusively monocausally, but models it as a complex system of different components. First, a systematic literature analysis is used to identify and describe factors that play a key role in the context of re-identification attacks and corresponding risk assessments. This is followed by a multi-step modelling process; the factors identified are first abstracted and combined into generalized factor groups, which are subsequently referred to as perspectives. In this context, criteria are presented for each factor that enable the characteristics of the respective component to be classified. Entity-relationship models are then used to illustrate the interactions that exist between the individual factors. In the final step, the interactions identified are used to construct a so-called influence matrix, which allows a quantitative description of the direction and characteristics of the relationships existing between the individual factors. Based on this matrix, various indices are presented that can be used to classify and stratify the re-identification risk of a data collection to be assessed and as a basis for targeted security measures.
Finally, two thought experiments are described, the aim of which is to assess the quality of the heuristics. Using published, verified re-identification attacks, the plausibility and selectivity of the risk stratification is confirmed; however, in this context, a somewhat lower assessment of the risk is shown than it probably actually was. Further research efforts are therefore necessary.2025-04-0
Analyse des Einflusses unterschiedlicher Sprunggelenksorthesenkonfigurationen auf die muskuläre Aktivität zur Fortbewegung
No full textSprunggelenksorthesen (engl. ankle foot orthosis; AFO) können helfen den menschlichen Gang zu unterstützen. In dieser Dissertation wurde der Effekt verschiedener Konfigurationen einer AFO mit unterschiedlichen Federsteifigkeiten (5, 10, 15, 20 und 25 kN/m) auf die muskuläre Aktivität der Plantarflexormuskulatur (M. Gastrocnemius; GAS und M. Soleus; SOL) und deren Antagonist (M. Tibialis anterior; TA) mit Hilfe von OpenSim simuliert und analysiert. Anschließend wurden die Ergebnisse der Simulation durch ein Experiment mit acht körperlich unbeeinträchtigten Probanden in der Realität überprüft. Hierfür wurde ein AFO mit drei verschiedenen Abständen zur Rotationsachse des Sprunggelenkes und fünf verschiedenen Federsteifigkeiten selber konfiguriert. Es wurde der Gang der Probanden mit den unterschiedlichen Orthesenkonfigurationen analysiert. Dabei wurde die Muskelaktivität des GAS, SOL und TA, der Sprunggelenkswinkel und Unterschenkelwinkel detektiert. In der Simulation konnte bei steigender Federsteifigkeit eine Reduktion der Muskelaktivität der Plantarflexoren um bis zu 65 % nachgewiesen werden. In unserem Experiment konnte kein signifikanter Einfluss unserer AFO auf die Muskelaktivität nachgewiesen werden. Jedoch zeigt sich durch das Tragen der AFO eine Veränderung des Sprunggelenkswinkels im Vergleich zur Kontrolluntersuchung ohne Orthese. Es zeigt sich jedoch eine hohe interindividuelle Variabilität der Ergebnisse, da sich bei einigen Probanden eine Reduktion, bei anderen jedoch auch ein Anstieg der Muskelaktivität zeigt. Ursächlich könnten hierfür Limitationen des Versuchsaufbaus oder des Modells unserer AFO sein. Die Ergebnisse dieser Arbeit legen nahe, dass eine Reduktion der Muskelaktivität durch eine AFO erreicht werden kann. Hierfür sind jedoch in der Zukunft weitere Experimente mit einer auf den Patienten individualisierten Anpassung einer AFO zu empfehlen. Mit diesen AFOs (aktiv oder passiv) sollten Untersuchungen mit verschiedenen Ganggeschwindigkeiten, Alltagaktivitäten bei gesunden Menschen und anschließend bei körperlich beeinträchtigten Personen durchgeführt werden.Ankle foot orthoses (AFOs) can help to support the human gait. In this dissertation, the effect of different configurations of an AFO with different spring stiffnesses (5, 10, 15, 20 and 25 kN/m) on the muscular activity of the plantarflexor muscles (gastrocnemius muscle; GAS and soleus muscle; SOL) and their antagonist (tibialis anterior muscle; TA) was simulated and analysed using OpenSim. The findings of the simulation were then subjected to investigation in a real-life experiment involving eight physically unimpaired test subjects. For this purpose, an AFO was configured with three different distances to the axis of rotation of the ankle joint and five different spring stiffnesses. The subjects' gaits were analysed with the different orthosis configurations. The muscle activity of the GAS, SOL and TA, the ankle joint angle and shank angle were detected. In the simulation study, a reduction in the muscle activity of the plantarflexor muscles of up to 65 % was demonstrated with increasing spring stiffness. In our experiment, no significant influence of our AFO on muscle activity could be detected. However, wearing the AFO resulted in a change in the ankle joint angle compared to the control case without the orthosis. However, there is a high degree of inter-individual variability in the results, as some subjects showed a reduction, while others showed an increase in muscle activity. Limitations of the test set-up or the AFO model could be responsible for this. The results of this study suggest that a reduction in muscle activity can be achieved with an AFO. However, further experiments with individualised adaptation of an AFO to the patient (passive or active AFOs) are recommended in the future. With these AFOs, studies should be carried out with different gait speeds, everyday activities in healthy people and then in physically impaired people.2025-03-2
Use of Click chemistry for neuronal protein staining in X10 expansion microscopy
No full textX10 Expansion Microscopy is a cost-effective method that allows for high-resolution imaging by expanding samples by a factor of ten through water absorption, enabling resolution down to 20 nm with conventional fluorescence microscopes. This technique can be combined with various staining methods for precise protein localization. However, traditional immunostaining methods suffer from a limitation known as the linkage error, where fluorophores are positioned up to 30 µm away from the target due to the size of antibodies.
Click chemistry offers a solution to this problem by enabling the direct, covalent binding of a fluorophore to the protein of interest (POI), eliminating the linkage error. This study investigates the use of click chemistry for localizing neuronal proteins in X10 expansion and compares it to traditional immunostaining. Hamster kidney cells (BHK), rat hippocampal neurons (HPC), and PC12 cells were transfected with constructs encoding unnatural amino acids for POI labeling via click chemistry. After expansion using X10, samples were imaged with conventional fluorescence microscopy.
The results show that click chemistry enables precise localization of proteins in X10 expansion, with PC12 cells demonstrating more accurate localization compared to traditional immunostaining. This study highlights the potential of combining click chemistry with X10 expansion to overcome the limitations of conventional staining methods, expanding the possibilities of super-resolution imaging.2025-04-1
Advancing Explainable AI in Medicine: Novel Frameworks for Robustness, Evaluation, and Human-Centered Interpretability
No full textThe advancement of artificial intelligence (AI) has led to its increasing usage in various
fields, including medicine, where AI models, particularly deep neural networks (DNNs),
have shown great performance. However, despite their predictive power, integrating AI
into clinical practice remains challenging due to transparency, ethical, and accountability
concerns. DNNs are often referred to as ªblack-boxº models because human users can
not understand their decision-making processes. This lack of transparency results in a lack
of trust, as clinicians are hesitant to rely on AI decisions without a clear understanding
of how these are made. Furthermore, regulatory frameworks such as the General Data
Protection Regulation (GDPR), the Health Insurance Portability and Accountability Act
(HIPAA), and the European Union AI Act mandate transparency and accountability in AI
decision-making, emphasising the need for explainability in AI systems.
Explainable AI (XAI) has emerged as a research area aimed at addressing these challenges
by providing insights into the decision-making of AI models, thereby increasing
interpretability and trust. However, despite the progress in XAI research, several issues remain
that prevent its integration into medicine. Some of these issues include (1) a lack of
robustness in XAI methods, where explanations can vary significantly due to small input
perturbations, hyperparameter adjustments, or differences in explanation methods (2) inconsistencies
in the evaluation of XAI methods, as there is no widely accepted standard for
evaluating explanations, leading to difficulties in comparing different XAI methods across
various data types and AI models and (3) the need for human-centered approaches that facilitate
interaction between AI developers and domain experts to ensure that explanations are
not only technically valid but also meaningful. To address these issues, this thesis presents
four main contributions.
First, a novel sample-wise rank normalisation technique is introduced to improve the
robustness of post-hoc XAI methods. This normalisation method reduces sensitivity to
small fluctuations in raw relevance attributions while preserving the relative importance
of features. By standardising XAI explanations, this method allows for more consistent
comparisons between different XAI methods across different data types and AI models.
Second, a comprehensive evaluation framework, BenchXAI, is developed to systematically
benchmark XAI methods across multiple biomedical data types, including clinical
data, medical imaging and signal data, and biomolecular data. Unlike previous evaluation
approaches that primarily rely on qualitative heatmaps or a limited set of evaluation
metrics, BenchXAI incorporates a Monte Carlo Cross Validation (MCCV) approach to ensure
that XAI evaluations are made across multiple randomised training and test splits.
This provides a more robust assessment of XAI methods. The benchmarking study conducted
using BenchXAI demonstrates that while certain XAI methods, such as DeepLift,
DeepLiftShap, GradientShap, and Integrated Gradients, perform consistently well across
various data types, others exhibit high variability, underscoring the need for standardisation
in XAI evaluation.
Third, this thesis proposes a novel ensemble majority vote approach to address the robustness
issue across different XAI methods. Individual XAI methods often introduce biases
based on their underlying explanation mechanisms, leading to inconsistent or contradictory
explanations. By aggregating explanations from multiple XAI methods using an ensemble
approach, this method reduces individual biases and enhances the stability of explanations.
This ensemble approach is applied to a large-scale electrocardiogram (ECG) classification
task, where it identifies clinically relevant features associated with different cardiovascular
diseases. Furthermore, the ensemble approach highlights significant differences in explanations
for true and false positive classifications, showing its potential for AI model debugging
and improving overall classification performance.
Finally, to bridge the gap between AI researchers and domain experts, this thesis introduces
CLARUS, an interactive XAI platform. CLARUS is tailored for graph neural
networks (GNNs) and allows domain experts to explore patient-specific protein-protein interaction
(PPI) networks, visualise XAI explanations, and interactively modify graph structures
to test counterfactual scenarios. This interactive approach enables domain experts
to improve their understanding of how AI models make decisions. By allowing users to
perform manual counterfactual analysis Ð such as investigating the effects of removing
or modifying specific gene interactions Ð CLARUS allows domain experts to validate AI
predictions. Unlike existing XAI platforms that primarily focus on static visualisations,
CLARUS allows for real-time retraining, making it a valuable tool for human-in-the-loop
AI evaluation.
Despite these advancements, challenges remain in further enhancing and expanding XAI
methods. Future research should focus on improving normalisation techniques to preserve
both robustness and magnitude in XAI explanations. Additionally, expanding the
BenchXAI framework to support more advanced AI models, such as attention-based models
and transformer networks, will be essential for better benchmarking studies. Furthermore,
usability studies on CLARUS are also necessary to assess its effectiveness in real-world
settings and optimise its design.
In conclusion, this thesis advances the field of XAI by addressing key limitations in robustness,
evaluation, and human-centeredness, ultimately contributing to the broader goal
of increasing trust in AI systems. By providing more stable and interpretable explanations,
developing standardised benchmarking frameworks, and fostering better collaboration between
AI researchers and domain experts, these contributions help bridge the gap between
AI research and clinical implementation.2025-06-3
Big Data Infrastructure for Analysing Digitalized Library Collections
No full textDigital Humanities (DH) represents an interdisciplinary field at the intersection of digital technologies and the study of the humanities. A variety of projects fall under the umbrella of DH, including digital archives, cultural analytics, online publishing, and other related endeavors. The present study focuses on text analysis in the context of DH. The objective of this work is to address two key challenges: the acquisition of data and the conduct of large-scale text analysis. The initial challenge arises from the difficulty in locating historical texts. The second issue arises from the fact that it is not a simple process for DH scientists to conduct an analysis on a large amount of text.
Following interviews with numerous DH scientists and discussions with relevant stakeholders, a list of functional and non-functional requirements has been compiled. In light of this, an evaluation of the available services on the market is conducted. It is regrettable that none of the aforementioned services aligns with our requirements. Consequently, a service has been developed with the objective of addressing the aforementioned issues. The newly developed service is designated as MINE. The service offers a search engine that enables users to locate historical texts from a range of data sources. Moreover, users are afforded the option of constructing corpora from the search results or uploaded files. Subsequently, users may instruct the system to analyze their corpora in accordance with the selected text analysis models and parameters. These analyses are executed on a high-performance cluster, which is a powerful computing infrastructure. This allows scientists to perform much larger analyses than they would be able to on their personal desktops or laptops.
Although MINE is still in the prototype phase at this time, the majority of the defined requirements have already been achieved. For features which are still under development or discussion, a comprehensive plan for their future implementation is also available.2025-08-0
Mikro-CT-Untersuchung zur Effektivität drei verschiedener Methoden für die Entfernung von Glasfaserstiften und die Beurteilung dadurch induzierter Rissbildung im Dentin
No full textThe aim of this in vitro study was to evaluate the effectiveness of three different methods for removing glass fiber posts and the resulting crack formation in dentin using micro-computed tomography. Effectiveness was measured by the working time, the loss of dentine, the induction of dentinal cracks, residual post and luting material, and procedural errors.
The study was conducted on forty-five extracted human maxillary anterior teeth with straight roots . Specimens were coronally shortened to a root length of 15 mm and divided into three groups (n = 15) based on canal volume (mm³) and canal surface area (mm³). All experimental steps were performed under a dental operating microscope at 10x magnification. Following a standardized preparation protocol, the root canals were shaped with NiTi instruments (Reciproc 50) and obturated in the apical 4 mm using warm vertical compaction. After a drying period of 2 hours, the samples were scanned preoperatively using micro-CT at 80 kV and 125 µA with a resolution of 5 µm. The preoperative scans served as the control group. After scanning, glass fiber posts (X-Post) were cut to a standardized length of 10 mm and adhesively luted into the root canals using a dual-cure resin cement. The posts were then removed using three different methods: Long-shaft round bur (EndoTracer #08), sonic-activated diamond tip (SonicFlex Endo tip #68), and DT Post Removal Kit. Working time was recorded with a stopwatch. After another drying time of 2 hours, a postoperative micro-CT scan was performed. The raw data from the scans were reconstructed, superimposed, and quantified using software programs. Results were then statistically analyzed and compared. Crack evaluation was performed using a scoring system that recorded both the course/origin and the number of newly formed cracks.
Removal with the SonicFlex tip resulted in the most dentin removal on average, with significant differences compared to the round bur and the DT Post Removal Kit. None of the three revision techniques could completely remove all post and luting material from the root canal, but the SonicFlex tip left the least residual material. Removal with the round bur took significantly the longest. Within the study, there were neither perforations nor instrument fractures. However, in all three test groups, deviations of the removal instrument from the tooth axis were observed (round bur: 2, SonicFlex tip: 1, DT Post Removal Kit: 2). These resulted in increased dentin removal and ledge formation. Each removal technique induced microcracks, with the DT Post Removal Kit causing the highest average number of microcracks and the SonicFlex tip the fewest. No correlation was found between dentin removal and crack formation.
Within the limitations of this study, all methods investigated for the removal of glass fiber posts induced microcracks, led to dentin loss, and left post and composite residues in the root canal. These results demonstrate that post removal, regardless of the technique used, carries a risk of iatrogenic defects. In light of current literature, the clinical relevance of studies on crack formation in dentin using extracted teeth should be further questioned in the future.2025-10-1
Einfluss der Funktionalisierung von Adhäsivsystemen mit polyedrischen oligomeren Silsesquioxanen auf die Dentinhaftung
No full textFor the adhesive bond between composite restorations and dentin is the formation of a hybrid layer through the application of an adhesive system essential. Over time, hydrolytic and enzymatic degradation processes lead to degeneration of the hybrid layer, reducing the longevity of the adhesive bond. Functionalized adhesive systems are intended to achieve increased resistance to degenerative degradation processes, e.g., by improving their mechanical properties or developing bioactive potential. Functionalization by adding MA-POSS-8 proved promising, as it improved the properties of adhesive systems and also demonstrated mineralization potential. However, there is a lack of research on how the addition of MA-POSS-8 affects the long-term adhesive strength of adhesive systems. Therefore, the present study investigated the shear strength of two commercially available adhesive systems (an etch-and-rinse adhesive and a universal adhesive in the self-etch process) that were functionalized with MA-POSS-8.
The adhesive systems Scotchbond Universal (3M, USA, universal adhesive used in the self-etch procedure) and Solobond Plus (VOCO GmbH, Germany, etch-and-rinse adhesive) were each modified with 5% and 10% MA POSS 8. The unmodified adhesives served as control groups. Bovine dentin test specimens were conditioned with the adhesives according to the manufacturer's instructions, and then a nanohybrid composite (Venus Diamond A3, Kulzer, Germany) was applied. The test specimens were artificially aged for 24 hours, 6 months, 12 months, or 24 months by water storage and subjected to a shear strength test followed by microscopic fracture analysis. Statistical evaluation was performed using one- and two-factor ANOVAs, Weibull analysis, and Chi2 tests (p < 0.05).
The initial bond strengths of the etch-and-rinse adhesive after 24 hours of storage showed no significant differences between the individual groups (control group: 17.4 ± 4.9 MPa; 5 wt% MA-POSS-8: 14.3 ± 4.3 MPa; 10 wt% MA POSS 8: 13.7 ± 4.4 MPa). After 12 months, the adhesive strength values of the etch-and-rinse adhesive mixed with 10 wt. % MA POSS 8 were significantly lower than in the control group (padj = 0.006). After 24 months of storage, the group with 5 wt.% MA-POSS-8 showed significantly higher adhesion values than the control group (padj. = 0.042) and the group with 10 wt.% MA-POSS-8 (padj. < 0.001) and furthermore led to more cohesive fracture patterns. Within the groups with the same MA-POSS-8 concentration, no significant differences were found between the different storage periods.
The initial bond strengths of the self-etch adhesive after 24 hours of storage were 19.1 ± 5.2 MPa in the control group, 20.0 ± 4.6 MPa in the group with 5 wt. % MA-POSS-8 and 18.6 ± 4.1 MPa in the group mixed with 10 wt. % MA-POSS-8. At no point was a difference between the adhesion values of the different groups observed. The shear strength of the self-etch adhesive increased in all groups over the storage period. The fracture patterns did not differ between the different groups.
The use of 5 wt% MA-POSS-8 does not impair the shear strength of dental adhesives but may improve the long-term stability of the adhesive bond.2025-09-1
Ultrafast exciton dynamics in moiré heterostructures
No full textDie optoelektronischen Eigenschaften von zweidimensionalen Halbleitern werden durch korrelierte Zwei-Teilchen Exzitonenzustände bestimmt, die durch die Coulomb-Wechselwirkung zwischen Ein-Teilchen-Löchern und Elektronen entstehen. Nach Anregung mit Licht werden Exzitonen in Übergangsmetall-Dichalkogenid-Monolagen und -Heterostrukturen von optisch hellen Zuständen in impulsindirekte dunkle Zustände gestreut. Diese sind nicht direkt zugänglich in rein optischer Spektroskopie.
In dieser Arbeit wird die ultraschnelle Dynamik der hellen und dunklen Zustände mittels zeitaufgelöster Impulsmikroskopie - einer Variante der zeit- und winkelaufgelösten Photoemissionsspektroskopie - untersucht. Ein Vergleich von Experiment und mikroskopischem Modell zeigt, dass Exzitonen-Phononen-Streuung der zugrundeliegende Mechanismus für die Bildung dunkler Exzitonen in WSe2-Monolagen ist. Darüber hinaus erfolgt die Bildung von Interlagen-Exzitonen (ILX) in einer gestaffelt ausgerichteten WSe2/MoS2-Bilage über einen kaskadenartigen Prozess, an dem hybridisierte Zustände beteiligt sind, wobei entweder das Elektron oder das Loch über die Grenzfläche transferiert. Das impulsaufgelöste Experiment ermöglicht die Analyse einer Signatur des Moiré-Übergitters, welches die Wechselwirkung des Exzitons mit dem Moiré-Potential visualisiert.The optoelectronic properties of two-dimensional semiconductors are defined by two-particle correlated exciton states formed via the Coulomb interaction between single-particle holes and electrons. After excitation with light, excitons in transition metal dichalcogenide monolayers and heterostructures undergo scattering from optically bright states to momentum-indirect dark states. The latter are not directly accessible in all-optical spectroscopies. In this thesis, the ultrafast dynamics of bright and dark states is probed by means of time-resolved momentum microscopy - a new variant of time- and angle-resolved photoemission spectroscopy. A comparison between experiment and microscopic modelling reveals exciton-phonon scattering as the underlying mechanism of momentum-indirect dark exciton formation in WSe2 monolayers. Furthermore, the formation of interlayer excitons (ILX) in a staggered band aligned WSe2/MoS2 bilayer occurs via a cascaded process involving hybridized states, in which either the electron or the hole component transfers across the interface. Notably, the momentum resolved experiments facilitate the direct analysis of a hallmark of the moiré superlattice, visualizing the interaction of excitons with the emerging moiré potential.2026-02-1
Surgical Aortic Valve Replacement in Patients Initially Evaluated for TAVI – Reasons for Conversion and Clinical Outcomes
No full textDiese Promotionsarbeit untersucht Patienten mit hochgradiger Aortenklappenstenose, die ursprünglich für eine TAVI evaluiert, letztlich jedoch einem AKE unterzogen wurden. Ziel war es, die Entscheidungsprozesse des interdisziplinären HT retrospektiv zu analysieren, die Gründe für den Prozedurwechsel zu identifizieren und die postoperativen Ergebnisse im Hinblick auf Mortalität und Morbidität zu bewerten. Dabei wurden sowohl die Relevanz präoperativer Risikostratifizierungsmodelle als auch die Übereinstimmung zwischen den geplanten und tatsächlich durchgeführten Eingriffen untersucht.
Die Ergebnisse zeigen, dass in 26,1 % (n = 36) der untersuchten Fälle, die ursprünglich von den zuweisenden Krankenhäusern für das TAVI-Verfahren vorgeschlagen wurden, das HT aufgrund eines als niedrig eingeschätzten Operationsrisikos eine konventionelle Operation empfahl. Diese Patienten wurden der sogenannten NER-Gruppe zugeordnet. In der KE-Gruppe (n = 57, 41,3%) erfolgte der Wechsel des Verfahrens primär aufgrund der Indikation zu Kombinationseingriffen, während in der ATE-Gruppe (n = 45, 32,6%) anatomisch-technische Limitationen den Wechsel des Verfahrens erforderlich machten.
Die gängigen Risikostratifizierungs-Scores, insbesondere der EuroSCORE I (additiv und logistisch) sowie der STS PROM-Score, korrelierten weitgehend mit der tatsächlich eingetretenen 30-Tage-Mortalität. Diese Korrelation ließ sich jedoch für den EuroSCORE II aufgrund einer höheren Zahl an Ausreißern nicht reproduzieren. Die Abweichung zwischen den präoperativ geplanten und den tatsächlich durchgeführten chirurgischen Verfahren war insgesamt gering und betraf lediglich 5,1 % der Fälle (n = 7).
Hinsichtlich der in-hospital Mortalität zeigte die KE-Gruppe mit 15,8 % die höchste Sterblichkeit, gefolgt von der ATE-Gruppe mit 15,6 %, während die Mortalität in der NER-Gruppe mit 5,6 % am niedrigsten war.
Auch die postoperative Morbidität variierte zwischen den Gruppen. Während in der KE-Gruppe die höchste Rate an postoperativen Komplikationen auftrat, darunter respiratorische und renale Komplikationen sowie verlängerte intensivmedizinische Aufenthalte, waren in der ATE-Gruppe ebenfalls relevante, aber weniger ausgeprägte Komplikationen zu verzeichnen. Die Patienten der NER-Gruppe wiesen insgesamt die geringste Morbidität auf, mit niedrigeren Raten an pulmonalen und renalen Komplikationen sowie kürzeren Intensivaufenthalten.
Die Ergebnisse dieser Arbeit zeigen zudem, dass die ursprüngliche Entscheidung zur TAVI-Zuweisung häufig auf einer primären klinischen Einschätzung beruhte, während die umfassendere Diagnostik durch das HT eine differenziertere Risikoabwägung ermöglichte. Dies verdeutlicht die essenzielle Rolle interdisziplinärer Entscheidungsfindung und multimodaler Diagnostik für die optimale Therapieplanung bei Patienten mit hochgradiger Aortenklappenstenose.This doctoral thesis examines patients with severe aortic valve stenosis who were initially evaluated for TAVI but ultimately underwent surgical aortic valve replacement (SAVR). The aim was to retrospectively analyze the decision-making processes of the interdisciplinary heart team (HT), identify the reasons for switching procedures, and assess postoperative outcomes with respect to mortality and morbidity. In addition, the relevance of preoperative risk stratification models and the concordance between planned and actually performed procedures were investigated.
The results show that in 26.1% (n = 36) of cases initially referred by external hospitals for TAVI, the HT recommended conventional surgery due to an assessed low surgical risk. These patients were assigned to the so-called NER group. In the KE group (n = 57, 41.3%), the procedure was switched primarily due to the indication for combined interventions, while in the ATE group (n = 45, 32.6%), anatomical or technical limitations necessitated the change.
Common risk stratification scores—particularly the EuroSCORE I (additive and logistic) and the STS PROM score—correlated largely with the observed 30-day mortality. However, this correlation could not be reproduced for the EuroSCORE II due to a higher number of outliers. The discrepancy between preoperatively planned and actually performed surgical procedures was low overall, affecting only 5.1% of cases (n = 7).
With respect to in-hospital mortality, the KE group showed the highest rate at 15.8%, followed by the ATE group at 15.6%, whereas mortality in the NER group was lowest at 5.6%.
Postoperative morbidity also varied between groups. The KE group demonstrated the highest rate of postoperative complications, including respiratory and renal events as well as prolonged ICU stays. Relevant but less pronounced complications were also observed in the ATE group. Patients in the NER group showed the lowest overall morbidity, with lower rates of pulmonary and renal complications and shorter ICU stays.
The findings of this thesis further demonstrate that the initial decision to refer patients for TAVI was often based on primary clinical judgment, whereas the more comprehensive diagnostic assessment by the HT allowed for a more differentiated risk evaluation. This underscores the essential role of interdisciplinary decision-making and multimodal diagnostics in optimizing therapy planning for patients with severe aortic valve stenosis.2025-10-0
Functional effects of the two transcription factors STAT1 and RUNX2 on the gene induction of interferon-responsive target genes in chondrogenic progenitor cells
No full textIm Zuge des Arthrose-bedingten Alterungsprozesses in menschlichem Kniegelenksknorpel wurde ein Reparaturweg nachgewiesen, bei dem chondrogene Progenitorzellen (CPC) als stammzellartige Vorläuferzellen zur Produktion von knorpelartigem Ersatzgewebe beitragen. Hierbei wurde für den Transkriptionsfaktor runt-related transcription factor 2 (RUNX2), auch bekannt unter den Bezeichnungen Osf2/Cbfa1, ein hemmender Einfluss in der Differenzierung von CPCs zu Chondroyzten und eine Begünstigung der osteoblastären Differenzierung nachgewiesen. Die nukleäre Translokation von RUNX2 wird durch Sequestrierung an zytoplasmatisch lokalisiertem, nicht-aktiviertem STAT1 (Signaltransduktor und Aktivator der Transkription 1) gehemmt. Unter inflammatorischen Bedingungen wie der Stimulation mit Zytokinen kommt es zu einer induzierbaren Phosphorylierung eines kritischen Tyrosinrestes im Carboxyterminus von STAT1 mit konsekutiver Akkumulation im Zellkern, wodurch die inhibitorische Wirkung auf die RUNX2-Translokation aufgehoben wird. Unklar ist allerdings, ob neben der STAT1-vermittelten Inhibition der nukleären RUNX2-Aktivierung umgekehrt auch eine Beeinflussung der STAT1-Signaltransduktion durch koexprimiertes bzw. defizientes RUNX2 stattfindet. In dieser Arbeit wurden deshalb chondrogene Progenitorzellen der Linie CPC241hT mit wildtypischer und fehlender RUNX2-Expression auf Unterschiede hinsichtlich ihrer Zytokin-induzierten, STAT1-vermittelten transkriptionellen Antwort untersucht. Die RUNX2-KO-Zellen zeigten in ihrer Kinetik der Interferon-stimulierten, nukleären Akkumulation keinen messbaren Unterschied zu den Wildtyp-Zellen. In Gelshift-Experimenten wurde eine Bindung von Tyrosin-phosphoryliertem STAT1 an eine hochaffine, spezifische Konsensus-Bindestelle, genannt γ-aktivierte Sequenz (GAS) in den Lysaten von Zytokin-exponierten Zellen nachgewiesen, die sich jedoch zwischen den CPC241hT-WT-Zellen und den CPC241hT-RUNX2 -/--Zellen nicht unterschieden. Für die hier untersuchten STAT1-responsiven Zielgene fand sich eine Stimulierbarkeit durch Interferon-α (IFN-α) bzw. IFN-γ in beiden Zelllinien, die sich damit als unabhängig von der Koexpression von RUNX2 erwies, jedoch Gen-spezifische Unterschiede erbrachte. Somit konnte mit den vorliegenden Experimenten kein sicherer Nachweis auf eine generelle Inhibition der STAT1-regulierten Signaltransduktion durch den osteogenen Transkriptionsfaktor RUNX2 dokumentiert werden. Zusammenfassend konnten keine Hinweise auf eine Beeinflussung des Transkriptionsfaktors STAT1 durch RUNX2-Expression nachgewiesen werden. Die aus der Literatur bekannte unidirektionale Inhibition von RUNX2 durch STAT1-vermittelte zytoplasmatische Retention konnte im umgekehrten Fall nicht beobachtet werden.In the context of the osteoarthritis-related ageing process in human knee joint cartilage, a repair pathway has been identified in which chondrogenic progenitor cells (CPCs) contribute as stem cell-like progenitor cells to the production of cartilage-like replacement tissue. The study revealed that the transcription factor RUNX2 (runt-related transcription factor 2), also known as Osf2/Cbfa1, exerts an inhibitory effect on the differentiation of CPCs into chondrocytes and promotes osteoblast differentiation. The nuclear translocation of RUNX2 is impeded by sequestration at cytoplasmically localised, non-activated STAT1 (signal transducer and activator of transcription 1). However, under inflammatory conditions, such as stimulation with cytokines, an inducible phosphorylation of a critical tyrosine residue in the carboxy terminus of STAT1 results in its accumulation in the cell nucleus, thereby preventing its inhibitory effect on RUNX2 translocation. Nevertheless, it remains unclear whether, in addition to the STAT1-mediated inhibition of nuclear RUNX2 activation, co-expressed or deleted RUNX2 also influences STAT1 signalling. Therefore, this study analysed differences in the cytokine-induced, STAT1-mediated transcriptional response of the CPC241hT line of progenitor cells, which expressed either wild-type or deficient RUNX2. The study revealed that the kinetics of interferon-stimulated nuclear accumulation were indistinguishable between RUNX2-KO cells and wild-type cells. In gelshift experiments, binding of tyrosine-phosphorylated STAT1 to a high-affinity consensus binding site called γ-activated sequence (GAS) was detected in lysates of cytokine-exposed cells, but did not differ between the CPC241hT-WT and CPC241hT-RUNX2 -/- cells. For the STAT1-responsive target genes investigated here, both cell lines could be stimulated by interferon-α (IFN-α) or IFN-γ, independent of the co-expression of RUNX2 and resulted in gene-specific differences. Thus, the present experiments did not provide reliable evidence for a general inhibition of STAT1-regulated signal transduction by the osteogenic transcription factor RUNX2. In summary, evidence of an influence of the transcription factor STAT1 by RUNX2 could not be demonstrated. The unidirectional inhibition of RUNX2 by STAT1-mediated cytoplasmic retention, known from the literature, could not be observed in the reverse case.2025-10-2