Institut für Informationsverarbeitung

Health services research faces the challenge of providing sound recommendations for action for the further development of health systems and care. The application of causal inference methods offers health services researchers an excellent opportunity to identify causal relationships under everyday conditions. The role of clinical trials with a classic randomised controlled trial (RCT) design is recognised as suitable for gaining insights that help establish causal inference, but other methodological approaches to generating evidence also play an important role in health services research.The discussion paper presents key concepts and assumptions of causal inference and highlights their relevance for health services research. The paper makes it clear that in order to fulfil the assumptions, it is necessary to integrate theory, contextual knowledge, understanding of mechanisms and formal concepts, such as directed acyclic graphs (DAGs), into a suitable empirical study design. To this end, RCTs, quasi-experimental methods, causal machine learning, target trial emulation, in silico trials and the mixed-methods approach of integrated inference are presented and discussed in terms of their applicability in health services research and their internal and external validity.All of the approaches presented here can contribute to the estimation of causal effects when used in a targeted manner and in accordance with the central assumptions. Their suitability depends largely on the research question, data quality, theoretical modelling and contextual knowledge. The combination of complementary designs and high quality data sources can increase the robustness of causal conclusions.Causal inference in health services research is not only a methodological procedure, but an integrative process that systematically combines theory, methodology and contextual knowledge. By consistently linking these aspects, health services research can generate differentiated and actionable insights that go beyond correlative analyses and enable an understanding of the mechanisms of causal processes. This can lead, for example, to evidence-supported recommendations that critically examine the often non-evidence-based status quo and reliably evaluate the benefits of new models.

SUMMARY & CONCLUSIONSMost, if not all, safety-critical systems developed are cyber-physical systems (CPS), defined by the presence of a physical component and connectedness to other systems, with a clear trend towards inclusion of over-the-air update capabilities. From a certification standpoint, every change to the system, even in software, should lead to an evaluation of its safety impact. Model-based system engineering offers the possibility of connecting domain-specific models used in the development of the CPS with relevant safety analysis models. The goal is to automate the evaluation of the impact of changes. Changes, however, can only be propagated if the dependencies of the specific model in which the change is introduced are accurately represented. These are difficult to model, as formal descriptions only exist for a subset of dependencies, and some have been completely neglected in model-based systems engineering. Our work provides an overview of the various shapes that dependencies take within model-based safety analysis. The importance of modeling these dependencies is shown in an illustrative example, in which an FMEA, SysML, and Lifecycle model are investigated.

Many countries try to shift their energy production to renewable energy sources in order to limit climate change and its negative effects. However, the success of renewable energy projects is dependent on important information, e.g., about areas proposed for such projects by local authorities or about available funding sources, that is usually spread across many sources of information. Consolidating all this information manually is a lot of work that must be performed regularly to stay up to date. We therefore propose an automated pipeline which monitors news, automatically detects duplicate articles from different sources, summarizes all relevant articles, and creates a single, georeferenced news feed. We leverage modern large language models (LLMs) and knowledge graphs for this purpose. The use of knowledge graphs future-proofs the news feed by enabling the future implementation of features such as tracking renewable energy projects across their lifetime. In our experiments we focus on wind energy in Germany, though our approach generalizes to other technologies and regions. We find that the web monitoring service Google Alerts is insufficient and that using Google Search instead leads to at least 164% additional relevant articles within our five-day test period. We achieve a precision of 41.6% at a recall of 98% for classification of articles into relevant or irrelevant to the topic of our news feed (wind and solar energy). Our approach for generating knowledge graphs from articles reaches an F 1 score of 89%, which improves upon an existing LLM-based baseline which reaches only 76%. Our entity resolution achieves an F1 score of 92% compared to a naïve name-based baseline of 66%.

Fixational eye movements are essential for visual perception but introduce significant challenges in ophthalmic imaging by causing motion artifacts that compromise the accuracy of diagnostic procedures. This study proposes an AI-based solution to achieve comprehensive motion correction in optical coherence tomography scans of the retina, with the aim of substantially enhancing image quality. In the context of intraoperative imaging, artifact correction is essential for accurate visualization, thereby mitigating the risk of diagnostic errors. A key challenge to the use of AI-based methods is the lack of extensive and appropriate training data, as no publicly available datasets currently contain paired optical coherence tomography images with and without motion artifacts. To bridge this gap, a simulation tool was developed to synthetically induce motion artifacts in OCT datasets. These artifacts are generated based on simulated fixational eye movements, creating realistic distortions that serve as a foundation for generating synthetic training data. Furthermore, existing AI models for motion extraction were evaluated and fine-tuned for the specific task of artifact recognition through transfer learning and subsequent artifact correction.

Conditional coding has lately emerged as the mainstream approach to learned video compression. However, a recent study shows that it may perform worse than residual coding when the information bottleneck arises. Conditional residual coding was thus proposed, creating a new school of thought to improve on conditional coding. Notably, conditional residual coding relies heavily on the assumption that the residual frame has a lower entropy rate than that of the intra frame. Recognizing that this assumption is not always true due to dis-occlusion phenomena or unreliable motion estimates, we propose a masked conditional residual coding scheme. It learns a soft mask to form a hybrid of conditional coding and conditional residual coding in a pixel adaptive manner. We introduce a Transformer-based conditional autoencoder. Several strategies are investigated with regard to how to condition a Transformer-based autoencoder for inter-frame coding, a topic that is largely under-explored. Additionally, we propose a channel transform module (CTM) to decorrelate the image latents along the channel dimension, with the aim of using the simple hyperprior to approach similar compression performance to the channel-wise autoregressive model. Experimental results confirm the superiority of our masked conditional residual transformer (termed MaskCRT) to both conditional coding and conditional residual coding. On commonly used datasets, MaskCRT shows comparable BD-rate results to VTM-17.0 under the low delay P configuration in terms of PSNR-RGB and outperforms VTM-17.0 in terms of MS-SSIM-RGB. It also opens up a new research direction for advancing learned video compression.

BackgroundChromosome organization plays an important role in biological processes such as replication, regulation, and transcription. One way to study the relationship between chromosome structure and its biological functions is through Hi-C studies, a genome-wide method for capturing chromosome conformation. Such studies generate vast amounts of data. The problem is exacerbated by the fact that chromosome organization is dynamic, requiring snapshots at different points in time, further increasing the amount of data to be stored. We present a novel approach called the High-Efficiency Contact Matrix Compressor (HiCMC) for efficient compression of Hi-C data.ResultsBy modeling the underlying structures found in the contact matrix, such as compartments and domains, HiCMC outperforms the state-of-the-art method CMC by approximately 8% and the other state-of-the-art methods cooler, LZMA, and bzip2 by over 50% across multiple cell lines and contact matrix resolutions. In addition, HiCMC integrates domain-specific information into the compressed bitstreams that it generates, and this information can be used to speed up downstream analyses.ConclusionHiCMC is a novel compression approach that utilizes intrinsic properties of contact matrix, such as compartments and domains. It allows for a better compression in comparison to the state-of-the-art methods. HiCMC is available at https://github.com/sXperfect/hicmc.

Reliable interaction force sensing between an exoskeleton and the human body is crucial for personalized support and efficient control. Poor donning and disturbances during movement lead to lower force measurement quality and thus to unfavorable use. In this work, we present an active smart human-robot interface by utilizing compliant actuator-sensor units (CASUs) that can be worn over clothing. These units combine capacitive force sensing and actuation using Shape-Memory Alloy (SMA) to actively compensate disturbances and improve force measurement by tightening the interface and reestablishing optimal contact. We present a pilot user study with five subjects, investigating the performance of an exoskele-ton interface featuring three integrated CASUs to compensate for non-ideal donning configurations and displacements during daily activities of walking, sitting down, and climbing the stairs. The results for a passive interface show a decrease in holding forces and force signal quality during activities. Our active interface, on the other hand, effectively compensates for disturbances, restoring high signal quality by the mean of peak-to-peak value up to 300% which leads to more distinct signals.

In the early stages of university education, frontal teaching within expansive lecture halls and paper-based assignments predominate. Students often encounter theoretical concepts whose practical relevance only emerges later, if at all. This can lead to reduced student motivation, an increased risk of academic disengagement, and a tendency toward superficial learning.Our newly developed first-semester course on digital circuits and systems employs an innovative approach that combines blended learning and challenge-based learning to address these issues effectively. Throughout the semester, we introduce four challenges, seamlessly integrated with the course lectures, designed to enhance students’ comprehension of the discussed topics. Each challenge presents a concise, well-defined task, tackled by small teams using tools such as circuit simulators, and our automated toolchain allows students to witness their circuit designs in action on FPGAs later.Through this challenge-based methodology, we aim to foster individual problem-solving skills and practical expertise, which we consider to be essential assets for students during their university education and future careers.

BackgroundBrain-derived neurotrophic factor (BDNF) is essential for antidepressant treatment of major depressive disorder (MDD). Our repeated studies suggest that DNA methylation of a specific CpG site in the promoter region of exon IV of the BDNF gene (CpG -87) might be predictive of the efficacy of monoaminergic antidepressants such as selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs), and others. This trial aims to evaluate whether knowing the biomarker is non-inferior to treatment-as-usual (TAU) regarding remission rates while exhibiting significantly fewer adverse events (AE).MethodsThe BDNF trial is a prospective, randomized, rater-blinded diagnostic study conducted at five university hospitals in Germany. The study’s main hypothesis is that {1} knowing the methylation status of CpG -87 is non-inferior to not knowing it with respect to the remission rate while it significantly reduces the AE rate in patients experiencing at least one AE. The baseline assessment will occur upon hospitalization and a follow-up assessment on day 49 (± 3). A telephone follow-up will be conducted on day 70 (± 3). A total of 256 patients will be recruited, and methylation will be evaluated in all participants. They will be randomly assigned to either the marker or the TAU group. In the marker group, the methylation results will be shared with both the patient and their treating physician. In the TAU group, neither the patients nor their treating physicians will receive the marker status. The primary endpoints include the rate of patients achieving remission on day 49 (± 3), defined as a score of ≤ 10 on the Hamilton Depression Rating Scale (HDRS-24), and the occurrence of AE.Ethics and disseminationThe trial protocol has received approval from the Institutional Review Boards at the five participating universities. This trial holds significance in generating valuable data on a predictive biomarker for antidepressant treatment in patients with MDD. The findings will be shared with study participants, disseminated through professional society meetings, and published in peer-reviewed journals.Trial registrationGerman Clinical Trial Register DRKS00032503. Registered on 17 August 2023.