An LLM-Based Visualization and Analysis Aid for a Secondary Use Clinical Data Platform.

This work introduces a novel approach to facilitate clinical research on secondary clinical data by integrating an LLM-based chatbot within a specialized platform called data hotel. The platform is designed to empower clinical researchers within our institution by enabling the generation of research hypotheses from secondary use patient data sources. Our focus in this work is on the deployment and functionality of the LLM-based chatbot within the data hotel ecosystem. The aim is to aid medical experts in visualizing and analyzing data sourced from the platform but also to enable the seamless storage of the generated code, enhancing the efficiency and reproducibility of the research process. This integration represents a significant advancement in leveraging LLM capabilities to enhance the utility and accessibility of clinical research platforms.

Continuous, Learned Imputation of Missing Values in Parkinson's Disease.

Parkinson's disease management requires accurate clinical scores but suffers from missing data. Leveraging self-supervised learning, we demonstrate superior generalization capabilities across populations compared to other well-established imputation techniques (MIWAE, MissForest, MICE). With the ability to employ the method already during the data collection and not afterward, the technology allows more robust data collection in clinical reality.

Learning debiased graph representations from the OMOP common data model for synthetic data generation.

BACKGROUND: Generating synthetic patient data is crucial for medical research, but common approaches build up on black-box models which do not allow for expert verification or intervention. We propose a highly available method which enables synthetic data generation from real patient records in a privacy preserving and compliant fashion, is interpretable and allows for expert intervention. METHODS: Our approach ties together two established tools in medical informatics, namely OMOP as a data standard for electronic health records and Synthea as a data synthetization method. For this study, data pipelines were built which extract data from OMOP, convert them into time series format, learn temporal rules by 2 statistical algorithms (Markov chain, TARM) and 3 algorithms of causal discovery (DYNOTEARS, J-PCMCI+, LiNGAM) and map the outputs into Synthea graphs. The graphs are evaluated quantitatively by their individual and relative complexity and qualitatively by medical experts. RESULTS: The algorithms were found to learn qualitatively and quantitatively different graph representations. Whereas the Markov chain results in extremely large graphs, TARM, DYNOTEARS, and J-PCMCI+ were found to reduce the data dimension during learning. The MultiGroupDirect LiNGAM algorithm was found to not be applicable to the problem statement at hand. CONCLUSION: Only TARM and DYNOTEARS are practical algorithms for real-world data in this use case. As causal discovery is a method to debias purely statistical relationships, the gradient-based causal discovery algorithm DYNOTEARS was found to be most suitable.

Improving Eye-Tracking Data Quality: A Framework for Reproducible Evaluation of Detection Algorithms.

High-quality eye-tracking data are crucial in behavioral sciences and medicine. Even with a solid understanding of the literature, selecting the most suitable algorithm for a specific research project poses a challenge. Empowering applied researchers to choose the best-fitting detector for their research needs is the primary contribution of this paper. We developed a framework to systematically assess and compare the effectiveness of 13 state-of-the-art algorithms through a unified application interface. Hence, we more than double the number of algorithms that are currently usable within a single software package and allow researchers to identify the best-suited algorithm for a given scientific setup. Our framework validation on retrospective data underscores its suitability for algorithm selection. Through a detailed and reproducible step-by-step workflow, we hope to contribute towards significantly improved data quality in scientific experiments.

A Unified Data Architecture for Assessing Motor Symptoms in Parkinson's Disease.

INTRODUCTION: The diagnosis and treatment of Parkinson's disease depend on the assessment of motor symptoms. Wearables and machine learning algorithms have emerged to collect large amounts of data and potentially support clinicians in clinical and ambulant settings. STATE OF THE ART: However, a systematical and reusable data architecture for storage, processing, and analysis of inertial sensor data is not available. Consequently, datasets vary significantly between studies and prevent comparability. CONCEPT: To simplify research on the neurodegenerative disorder, we propose an efficient and real-time-optimized architecture compatible with HL7 FHIR backed by a relational database schema. LESSONS LEARNED: We can verify the adequate performance of the system on an experimental benchmark and in a clinical experiment. However, existing standards need to be further optimized to be fully sufficient for data with high temporal resolution.

Understanding Human-Computer Interactions in Restricted Clinical Environments.

INTRODUCTION: Conducting research on human-computer interaction and information retrieval requires unobtrusive observations within existing network architectures. STATE OF THE ART: Most of the available tools are not suitable to be applied within restricted clinical systems. The specific requirements hinder analysis of the human factors in health sciences. CONCEPT: We identified extensions for popular web browsers as a suitable way to conduct studies in highly regulated environments. IMPLEMENTATION: Considering the specialized requirements and an adequate level of transparency for the recorded clinician, we developed an open-source Web Extension compatible with major web browsers. LESSONS LEARNED: We identified the challenges associated with the specific tool and are preparing its use to understand clinical reasoning in personalized oncology.

Mapping the Oncological Basis Dataset to the Standardized Vocabularies of a Common Data Model: A Feasibility Study.

In their joint effort against cancer, all involved parties within the German healthcare system are obligated to report diagnostics, treatments, progression, and follow-up information for tumor patients to the respective cancer registries. Given the federal structure of Germany, the oncological basis dataset (oBDS) operates as the legally required national standard for oncological reporting. Unfortunately, the usage of various documentation software solutions leads to semantic and technical heterogeneity of the data, complicating the establishment of research networks and collective data analysis. Within this feasibility study, we evaluated the transferability of all oBDS characteristics to the standardized vocabularies, a metadata repository of the observational medical outcomes partnership (OMOP) common data model (CDM). A total of 17,844 oBDS expressions were mapped automatically or manually to standardized concepts of the OMOP CDM. In a second step, we converted real patient data retrieved from the Hamburg Cancer Registry to the new terminologies. Given our pipeline, we transformed 1773.373 cancer-related data elements to the OMOP CDM. The mapping of the oBDS to the standardized vocabularies of the OMOP CDM promotes the semantic interoperability of oncological data in Germany. Moreover, it allows the participation in network studies of the observational health data sciences and informatics under the usage of federated analysis beyond the level of individual countries.

Development of an immediate release excipient composition for 3D printing via direct powder extrusion in a hospital.

3D printing offers the possibility to prepare personalized tablets on demand, making it an intriguing technology for hospital pharmacies. For the implementation of 3D-printed tablets into the digital Closed Loop Medication Management system, the required tablet formulation and development of the manufacturing process as well as the pharmaceutical validation were conducted. The goal of the formulation development was to enable an optimal printing process and rapid dissolution of the printed tablets for the selected model drugs Levodopa/Carbidopa. The 3D printed tablets were prepared by direct powder extrusion. Printability, thermal properties, disintegration, dissolution, physical properties and storage stability were investigated by employing analytical methods such as HPLC-UV, DSC and TGA. The developed formulation shows a high dose accuracy and an immediate drug release for Levodopa. In addition, the tablets exhibit high crushing strength and very low friability. Unfortunately, Carbidopa did not tolerate the printing process. This is the first study to develop an immediate release excipient composition via direct powder extrusion in a hospital pharmacy setting. The developed process is suitable for the implementation in Closed-Loop Medication Management systems in hospital pharmacies and could therefore contribute to medication safety.

Visual Analytics of Gaze Data with Standard Multimedia Players.

With the increasing number of studies, where participants' eye movements are tracked while watching videos, the volume of gaze data records is growing tremendously. Unfortunately, in most cases, such data are collected in separate files in custom-made or proprietary data formats. These data are difficult to access even for experts and effectively inaccessible for non-experts. Normally expensive or custom-made software is necessary for their analysis. We address this problem by using existing multimedia container formats for distributing and archiving eye-tracking and gaze data bundled with the stimuli data. We define an exchange format that can be interpreted by standard multimedia players and can be streamed via the Internet. We convert several gaze data sets into our format, demonstrating the feasibility of our approach and allowing to visualize these data with standard multimedia players. We also introduce two VLC player add-ons, allowing for further visual analytics. We discuss the benefit of gaze data in a multimedia container and explain possible visual analytics approaches based on our implementations, converted datasets, and first user interviews.