Schwer­punkte

The work of the project team focuses on the following areas, which together form the foundation of a connected innovation ecosystem for Artificial Intelligence in medicine:

The KIMed network establishes sustainable governance structures to ensure long-term, reliable and transparent collaboration between all participating stakeholders. At its core is the development of a sustainable organisational structure that is intended, in the future, to lead to the establishment of an appropriate legal entity. The aim is to secure collaboration beyond the project duration and to create stable conditions for cooperation, innovation, and growth.

A key component is the coordinated collaboration between clinical institutions, scientific partners, industry stakeholders, and political and regulatory institutions. The network assumes a structuring role by integrating different perspectives, developing shared visions and supporting the implementation of concrete initiatives. This ensures that activities are aligned and synergies are utilised.

At the same time, the network is continuously expanded, maintained and strategically developed. This includes the targeted expansion of the partner network as well as the further refinement of shared topics and priorities. The focus is on creating a high-performance innovation ecosystem for AI in medicine that brings together regional strengths while remaining internationally connected.

In addition, the network organises various exchange formats such as the Network Partner Forum, webinars and interactive sessions. These promote networking, professional exchange and the initiation of joint projects.

Secure Processing Environments (SPEs) are controlled, technically secured working environments in which sensitive data – particularly health data – can be processed without leaving the protected environment. They therefore represent a key approach to ensuring data protection, data security and, at the same time, the usability of data for research and development.

Within KIMed, the focus is not on building such an environment, but on developing a robust concept for an SPE. The aim is to systematically identify the technical, regulatory and organisational requirements necessary for the operation of such an environment.

Particular attention is given to the conditions under which health data can be processed within an SPE and AI algorithms can be applied securely to these data. In addition to technical aspects such as access control, infrastructure and data integration, legal and regulatory frameworks as well as requirements relating to governance, use and operation are also considered.

The concept is based on a structured catalogue of requirements developed in coordination with academic and industry partners. Different operational and implementation models are also analysed in order to represent various scenarios for the deployment of an SPE.

The result is a transferable, practice-oriented concept that can serve as a foundation for the implementation of secure processing environments for health data and AI applications – both within the KIMed network and beyond.

Within KIMed, a portal is being developed that will serve as a central information platform for health data and, in the future, also for AI algorithms.

The aim is to improve the discoverability and transparent description of health data. To achieve this, available data sources are recorded using a standardised metadata format suitable for medical applications. This structured description enables key dataset characteristics, such as content, data type, quality, collection context and usage context, to be presented in a transparent and understandable way.

On this basis, the KIMed Portal supports users in searching for relevant health data and assessing whether these data are suitable for specific research or development projects.

In addition, an overview of AI algorithms in the context of health data is planned. The focus is on methods that have already been applied in concrete use cases. Through the structured description of these algorithms and their application contexts, reusability and transferability to new research questions will be facilitated.

The development of the portal takes existing approaches and initiatives into account. At the same time, implementation is aligned with current national and European developments, particularly in the context of standardised metadata structures for health data.

The KIMed Portal therefore provides a foundation for systematically discovering health data and supporting their use in AI-based applications.

This focus area introduces exemplary use cases and analysis pipelines as demonstrators within the Secure Processing Environment (SPE). The aim is to test and evaluate the functionality of the infrastructure, the network services provided and the underlying infrastructure components in a practical setting. Existing tools and methods are used as software demonstrators in the form of proof-of-concept implementations. On this basis, the technical and organisational framework conditions (governance rules) developed within the project are implemented and assessed for their practical use within the protected runtime environment.

Demonstrator scenarios currently under consideration include:

• Evaluation of text analysis pipelines through the integration of existing workflows (e.g. from the GeMTeX project) for the de-identification and annotation of clinical texts as a data-driven text-processing scenario.
• Integration and technical feasibility assessment of domain-specific tools, such as computer-assisted phenotyping frameworks (e.g. the TOP pipeline) or OMICS analysis tools.
• Validation of image-processing workflows through the integration of existing tools for AI-based image anonymisation (such as defacing algorithms for head MRI scans with automated quality control) in order to test the processing of large 3D volumetric datasets within the SPE.
• Assessment of multimodal data integration through the exemplary incorporation of processing pipelines that link heterogeneous data sources (e.g. environmental exposome data with phenotypicn LIFE cohort data), in order to evaluate the performance and structuring capabilities of the SPE infrastructure for complex mental health research questions.

Through the technical demonstration of these exemplary pipelines, the concrete added value of the Saxon SPE services – particularly with regard to the reusability of algorithms and privacy-compliant data sharing – is made transparent and measurable for the biomedical research community.

The KIMed network provides its partners with a structured platform for targeted training opportunities and professional exchange, particularly through flexible online formats. The focus is on needs-based support for content-related and methodological questions concerning medical data processing and the use of AI. To this end, the network facilitates access to experts, promotes networking among partners and simplifies access to specialised knowledge.

A key component is the provision of tailored training opportunities that are directly aligned with the requirements of network partners. The aim is to systematically strengthen competencies in the secure and privacy-compliant handling of medical data and AI applications. At the same time, efforts are made to establish a common level of knowledge across the network in order to improve interdisciplinary collaboration and reduce communication barriers between different fields of expertise.

The quality and relevance of these activities are ensured through continuous feedback from network partners and ongoing optimisation. Training courses, information events and accompanying exchange formats are regularly evaluated and further developed. In this way, a dynamic learning and exchange environment is created that provides different target groups with knowledge relating to AI in medicine while actively contributing to the further development of the network’s structures and content.

Communication within KIMed makes the network’s diverse activities visible, understandable and accessible. The target groups are broad and include the scientific community from universities, universities of applied sciences and research institutes, physicians, clinician scientists and healthcare institutions, as well as industry stakeholders ranging from established companies to start-ups in the fields of AI development and medical technology. In addition, communication activities address policymakers, professional associations, funding organisations and the interested public.

A central task of communication is to clearly convey the importance and concrete benefits of KIMed, attract new partners to the network and strengthen trust in the responsible handling of sensitive health data. At the same time, it positions Saxony as an innovative location for medical AI. These key messages are communicated in clear, scientifically sound language that addresses both experts and non-specialist audiences.

To achieve this, the communication team combines digital and traditional communication formats. These include, among others, the website, LinkedIn, scientific publications, events and workshops, as well as traditional media relations and participation in trade fairs. A consistent visual identity ensures strong recognition and visibility.