One of the most important goals of the Medical Informatics Initiative is to promote young scientists. Well-trained specialists play a central role in the future of digitization in medicine. Specialized “data scientists” should both master the latest tools of informatics and be familiar with medical terminology so that they can meaningfully evaluate the growing volumes of data. To establish medical informatics as a progressive field in research and education, junior research groups support professorships at various SMITH sites. The junior research groups of the SMITH Consortium develop forward-looking perspectives on medical informatics issues and future scientists.
In the junior research group “Benefit Optimization and Availability Enhancement of Health Data (BENEFIT)”, young scientists under the leadership of Dr. Sylvia Nürnberg are working on the secondary use of health data.
Through the Medical Informatics Initiative (MII), data from university hospitals are made available in Data Integration Centers across sites for research and healthcare. The junior research group BENEFIT develops new concepts and methods to improve research opportunities and patient care with the help of the data. For this purpose, the group investigates who future users would be and what tasks would they like to perform on the data. From this, the group will develop and identify the most relevant data sets along applicable ethical, legal and data protection guidelines.
Especially in the field of artificial intelligence and its application to medical devices, an exciting field of topics is developing. Among other things, the new EU regulation for artificial intelligence handles intellectual property and AI, where the patent law principle of the creating or inventing person does not apply. The Medical Devices Regulation also poses major challenges for a practical implementation of medical data from patient care.
Principle business models can be derived from the work of the junior research group, which can increasingly convince university data owners of the advantages of making data available.
Management: Dr.-Ing. Sasanka Potluri
Funding period: 2020 – 2025
University Hospital Jena
Institute for Medical Statistics, Informatics and Data Science (IMSID)
The establishment of the junior research group aims to improve patient care through artificial intelligence applications. This is accomplished by supporting demand planning and the structuring of processes through more accurate demand prediction and simulations of clinical processes. To this end, machine learning techniques from healthcare data are combined with methods of mathematical optimization and simulation. The focus of this group is the development, use and comparative evaluation of Deep Learning for the analysis and update of incomplete multivariate time series from clinical care processes.
An indispensable prerequisite of the project is that data from patient care can be incorporated into the aforementioned methods. To this end, the junior research group will develop approaches for the acquisition, pooling and quality assurance of anonymous training and test data in close collaboration with the local Data Integration Center. The junior research group will develop and evaluate deployable prototypes for selected tasks, which can be the starting point for downstream product development. The results will allow an assessment of possible efficiency gains, evaluate the routine capability and stability of the methods and, if necessary, consider possible side effects of their use. A particular concern here is to prevent undesirable developments that could result from algorithms that are opaque, cannot be scrutinized and only evaluate a section of reality. Instead, the project aims at the comprehensibility and criticality of the approaches used. Special attention is given to supporting the calibration of data-driven applications. Through calibration, data-driven methods must account for the fact that training data have origin-specific characteristics, for example, due to the population from which the data originate.
Head: Dr. Ivana Kraiselburd
Funding period: 2022 – 2026
University Hospital Essen
Institute for Artificial Intelligence in Medicine (IKIM)
Chair for Data Science
The junior research group “SepsisPrep” is led by Dr. Ivana Kraiselburd and belongs to the Institute for Artificial Intelligence at the University Medical Center Essen. The group is in the Department of Data Science.
“SepsisPrep” researches methods with which microbiome data from sepsis patients can be processed, evaluated and integrated. This primarily involves DNA sequence data obtained by sequencing gut and skin flora. These data can be used to predict the onset of sepsis as well as possible antibiotic resistance, which can lead to problems in the treatment of sepsis.
A personalized microbiome profile will be created for individual patients based on the sequence data and this will be used to create a model for early prediction of sepsis. Such a model can complement existing sepsis prediction models based on clinical data and offers the possibility to accelerate the prediction of sepsis.
The group is specifically dedicated to the study of sepsis in critical care patients. Sequencing and analysis of microbiome data will be used to search for molecular signatures suitable for early detection of sepsis or characterization of sepsis risk. In addition, the junior research group aims to establish tools for the detection of antibiotic resistance profiles in the gut and skin microbiome in patients. In this way, this project will complement existing methods for predicting the evolution of disease-causing pathogens as well as their potential antibiotic resistance, and will be able to expand the predictive horizon.
Head: Dr. Alexandr Uciteli
Funding period: 2021 – 2024
Faculty of Medicine
Institute of Medical Informatics, Statistics and Epidemiology (IMISE)
The goal of the project is to develop and apply an ontology-based software platform for complex phenotypes (TOP Framework). The framework will build on the medical data available at the Data Integration Centers within the Medical Informatics Initiative. The concept will be developed along with the use cases that are a focus of the SMITH consortium, but will not remain limited to these and should later be applicable to many use cases of comparable projects in the medical domain. To this end, the concept will be implemented in a modular web application that combines various software tools and services for algorithmic phenotyping. The developed phenotype algorithms including further metadata will be made available in standardized (and if possible public) digital archives (repositories).
The TOP framework will consist of three main components. The Structured Data Module will support the modelling and execution of phenotype algorithms and phenotypic queries (inclusion/exclusion criteria, feasibility queries, cohort formation, data sharing) as well as advanced analyses on structured data. The Text Data Module will be linked to an existing search engine where relevant documents (e.g., physician letters) will be indexed to enable searching and structuring of relevant clinical information in text. The phenotype algorithms and other metadata will be managed in digital archives to be developed and made available via a web portal and standardized interfaces.
The framework in modelling phenotypes will support clinical experts in identifying relevant individuals for further analyses or studies, as well as selecting and analysing relevant data.