Clinical Decision Support
Clinical decision support (CDS) provides physicians, healthcare professionals, patients or other individuals with knowledge and person-specific information that is intelligently filtered or presented at the right time to improve health and healthcare.
Apache Superset meets Kafka and Deep-Learning
Digital processes only make sense if the resulting data can also be analyzed. This requires powerful and flexible visualization options. The aim of the project is to establish a modern dashboard infrastructure that can display the current status of the university hospital and predict this into the future using AI methods.
Kontakt: Mohammad Fattouh, Torben Lauck
Advancing Cardiovascular Risk Identification with Structured Clinical Documentation and Biosignal Derived Phenotypes Synthesis
The aim of the BMBF consortium project ACRIBiS is to improve individual risk assessment in the prevention, diagnosis and treatment of cardiovascular diseases. To this end, clinical information is being combined with biosignal data at 15 partner sites and the standardized processing of raw ECG data from around 4,500 patients is being investigated. The establishment of the biosignal analyses is based on the infrastructure of the Medical Informatics Initiative (MII) and aims to include patients in the individualized risk assessment by providing an app. ACRIBIS can thus make an important contribution to the development of an interoperable medical data infrastructure and pave the way for a demonstrably effective and dynamically adaptive clinical decision-making aid at patient level. The aim of the sub-project in Freiburg is to implement the concepts developed in the consortium for the combination of biosignal processing with structured data and the recruitment of around 250 patients. To this end, the ACRIBiS infrastructure will be integrated into routine processes and the Data Integration Center (DIZ).
Duration: 2023 - 2027
Sponsor: Implementierung der Biosignal-Phänotypisierung (gesundheitsforschung-bmbf.de)
Contact: Sebastian Siegel
INTERventional POLypharmacy – drug interActions – Risks
The aim of the BMBF consortium project INTERPOLAR is to use an algorithm to identify patients in hospitals with a particularly high risk of medication errors and side effects. Once these are known, ward pharmacists can direct their resources precisely to where maximum patient benefit can be achieved. The results of the intervention will be examined by means of a controlled randomized study. Proof of the benefits of the intervention will significantly promote the introduction of the model in other hospitals.
A total of 17 universities and university hospitals are involved in this medical informatics use case of the Medical Informatics Initiative (MII ). The aim of the sub-project in Freiburg is to conduct a translational study to investigate whether the algorithms developed also deliver similar results in other locations in order to test whether the results are robust and generalizable. To this end, the intervention will be used on six different wards at the university hospital.
Duration: 2023 - 2026
Sponsor: INTERPOLAR - Medizininformatik-Use Case "INTERventional POLypharmacy – drug interActions – Risks" (gesundheitsforschung-bmbf.de)
Project partner: Apotheke der Uniklinik Freiburg
Contact: Torben Lauck
Reducing the risk of delirium with AI
Together with care professionals, the researchers in the KIDELIR project are developing a hybrid AI support system for detecting the risk of delirium. The aim is to create a practical decision-making aid for care professionals that supports the implementation of timely and individualized prevention and treatment measures in day-to-day care. The underlying database is made up of data from different sources, which is merged and enables the system to provide appropriate recommendations for action by recognizing patterns. At the Institute for Digitalization in Medicine (IDIM), the delirium-relevant data is extracted from the electronic patient file and made available for real-time AI analyses.
Duration: 2022 - 2025
Sponsor: KIDELIR — Miteinander durch Innovation (interaktive-technologien.de)
Project partners: Geriatrie der Uniklinik Freiburg, Hochschule Furtwangen, Meona GmbH Freiburg
Contact: Mohammad Fattouh
The digital nervous system of the hospital information system (HIS)
The monitoring of clinical processes with regard to quality and adherence to guidelines as well as their control through Clinical Decision Support (CDS) can make an important contribution to increasing treatment safety. This requires the integration of all decision-relevant information, such as the results of diagnostic imaging, pathology findings, clinical findings and medical history. This information is currently distributed across various IT systems and/or only available as free text. The UKF is therefore planning to merge the data via event streaming with a central FHIR server as a clinical data repository and make it usable for CDS.
Duration: 2021 - 2025
Sponsor: Krankenhauszukunftsgesetz (KHZG) - BMG (bundesgesundheitsministerium.de)
Contact: Christian Haverkamp
Translational Projects
The aim of our translational projects is to develop new digital solutions in the healthcare sector and innovative digitalization concepts for the hospital.
AI in the neurosurgical operating room
The FRAI.lab in the neurosurgical operating room is a project of the KI-Allianz BW and concretizes the intersection of medical technology, AI and precision medicine. Specific use cases for the real-world testing of complex medical products are being developed and implemented with various partners. The aim is to integrate them into the real workflow of neurosurgery and the hospital information system of the clinic as well as to develop multidimensional sensor data for AI applications. The commercial interests of all partners involved will be taken into account so that a long-term financing model can be developed.
Duration: 2024 - 2025
Sponsor: Ministerium für Wirtschaft, Arbeit und Tourismus Baden-Württemberg, Mertelsmann Foundation
Project partners: Klinik für Neurochirurgie, Uniklinik Freiburg
Contact: Felix Heilmeyer
In the context of the KI-Allianz BW, AI innovation clusters are being established and networked in various regions of the state of Baden-Württemberg. One project is concerned with the development of a comprehensive data platform that will provide stakeholders with the easiest possible access to high-quality data sets that are essential for the development and application of reliable AI solutions. Medical data poses particular challenges in terms of data protection, IT security and ethics.
Five work packages are being implemented at the UKF: (1) The medical data developed for research purposes as part of the Medical Informatics Initiative and the BW Health Cloud will be made available for the common good. (2) Synthetic data will be generated that can be used for non-medical applications. (3) A concept is developed for use cases where it is not possible to sufficiently anonymize the data (bring the algorithms to the data). (4) A contract template is created that regulates the training of AI models on medical data. (5) A legal and ethical conformity assessment of the solutions developed in the consortium is carried out.
Duration: 2024 - 2025
Sponsor: Ministerium für Wirtschaft, Arbeit und Tourismus Baden-Württemberg
Project partners: Fraunhofer-Institute IOSB & IPA, KIT & FZI Karlsruhe, Universitäten Stuttgart & Tübingen, Stadt Freiburg, Hochschule Aalen
Contact: Martin Hinze
In cooperation with the Department of Neurology, we are investigating whether commercial systems for appointment scheduling, such as Doctolib, are suitable for mapping the workflow of university outpatient clinics with their complex requirements.
Contact: Christian Haverkamp
Currently, there are manufacturer-dependent forms in the various hospital information system tools that are not interoperable. FHIR (Fast Healthcare Interoperability Resources) questionnaires make it possible to store, exchange and process structured information via questionnaires, queries and assessments. They define how this information should be structured and coded to ensure uniform presentation and interpretation.
Contact: Christian Haverkamp
In this cooperation project with Averbis, we will explore and develop the possible applications of large language models (LLMs) in the healthcare sector. The aim is to create innovative solutions that optimize the documentary process in the clinical context. By combining the medical expertise and patient data of the University Medical Center Freiburg with the expertise of Averbis in the field of AI-supported text analysis and processing, tailor-made models can be developed that meet the complex requirements of the clinical work environment.
Contact: Christian Haverkamp, Felix Heilmeyer
Sensor-assisted diagnostics
In sensor-assisted diagnostics, data is continuously collected and analyzed to monitor vital signs and other important health indicators.
Regular rounds by doctors and nurses to check on the patient's current condition are an important part of the ICU routine. However, these only reflect the patient's condition at a particular point in time. Some brief and initially seemingly insignificant movements during the rest of the day could be overlooked. However, it is precisely these movement patterns that could provide crucial information and have great significance for the success of the treatment. To this end, we want to develop a system that is able to continuously monitor the motor activity of intensive care patients. To achieve this, we will use various image analysis options, e.g. the recording of a thermal imaging camera.
Contact: Torben Lauck
The precise and rapid detection of shockable cardiac conditions is a decisive step towards improving the chances of survival of patients with sudden cardiac arrest. The aim of the project is to investigate the automated feature extraction and classification of ECG data with residual neural networks. The aim is to improve the algorithms used by automatic external defibrillators to recognize shockable rhythms.
Contact: sebastian Siegel
The early treatment of a stroke is crucial for minimizing brain damage. Therefore, methods are being sought to detect stroke symptoms as early as possible, e.g. via the "Face, Arms, and Speech Test" (FAST). This project is investigating whether the arm weakness test can be carried out with the integrated video tracking of an iPhone.
Contact: Vittorio Lippi
Innovative EXOskeleton control system for SMOOTH Assistance
Exoskeletons are wearable robotic devices that are placed around a person's body to support, enhance or enable movement. This technology, which often consists of a combination of mechanical elements, sensors and actuators, can be used in various applications, from medical rehabilitation to military use. The EXOSMOOTH project is investigating the effectiveness of a novel control strategy for fluid support, where the phase transitions are controlled so that the exoskeletons provide smoother support to the user.
Contact: Vittorio Lippi
Further information: Posturob II
Coded Infrastructure on Kubernetes
Cloud technologies enable a new dimension of software development, but require a different strategy in the design of the infrastructure. The aim of the project is to establish a coded infrastructure based on Kubernetes with the help of a modern DevOps strategy. This means that the state of the infrastructure is no longer determined by manual settings, but by auditable and versionable source code. Complex server architectures can be recreated or copied for testing purposes in just a few minutes.
Contact: Felix Heilmeyer