Data and AI Readiness

Successful medical AI development starts with the right data foundation. The Research Cluster ‘Data and AI Readiness’ combines technologies, processes, and expertise to transform complex real-world medical data into impactful AI solutions.

Fraunhofer MEVIS addresses the critical challenges that determine whether AI succeeds in clinical practice: collecting, curating, and annotating medical data with rigorous quality assurance; ensuring datasets are representative; integrating heterogeneous or longitudinal data sources; navigating regulatory requirements (FDA, MDR, EU AI Act); building efficient development workflows—from streamlined training pipelines to foundation model adaptation; and thoroughly evaluating AI systems for robustness and reliability.

The experts of Fraunhofer MEVIS develop tailored solutions for each use case, enabling medical AI projects to progress confidently from research to real-world impact.

Data Collection, Curation, and Annotation 

Fraunhofer MEVIS develops advanced data preprocessing solutions including efficient annotation technology, automated label extraction using large language models (LLMs), synthetic data generation and augmentation, exploratory analysis and harmonization of large datasets, anomaly detection, and data quality assurance. This represents the foundation of successful AI development.

Interoperability and Data Integration 

These are essential factors to advance medical knowledge and improve healthcare delivery. Effective integration of relevant patient history from different data sources and available clinical information enables comprehensive analyses, gain of medical knowledge and improved patient care. Fraunhofer MEVIS  has extensive expertise in managing diverse data structures, including DICOM, FHIR, and OMOP and is a key contributor to various national data infrastructure projects (e.g. NAKO, RACOON, NFDI4Health, KI-FDZ, CanConnect).

AI Evaluation & Monitoring

AI evaluation and monitoring are indispensable for the development and governance of medical AI, enabling the reliable clinical deployment of AI solutions. Fraunhofer MEVIS develops rigorous evaluation and monitoring approaches that assess AI system robustness, fairness, and generalizability across patient sub-cohorts, technical variability, and dataset shifts. An example of such monitoring is implemented for mammography screening models in the EURAMET project MAIBAI.

Regulatory Compliance

Compliance with evolving regulations (MDR, FDA, AI Act, and others) is essential for medical AI development. Fraunhofer MEVIS leverages traceable development processes, structured data management, quality management systems, and rigorous evaluation and documentation to deliver software solutions ready for regulatory submission and approval. EN ISO 13485-certified since 2005, it has over two decades of experience supporting regulatory compliance, as demonstrated in EU-funded projects such as ODELIA and COMFORT.

• End-to-end machine learning platform for tabular data: Graphical prototyping toolbox for ML experimentation repository
• ETL: From the German Health Data Lab data formats to the OMOP Common Data Model. Finster et al. 2025. Article
• Common data models and data standards for tabular health data: a systematic review. Finster et al. 2025. Article
• meval: A Statistical Toolbox for Fine-Grained Model Performance Analysis. Sutariya & Petersen, 2025. Article, Toolbox
• Combining Machine Learning With Real-World Data to Identify Gaps in Clinical Practice Guidelines: Feasibility Study Using the Prospective German Stroke Registry and the National Acute Ischemic Stroke Guidelines. Müller et al. 2025. Article
• Slicing Through Bias: Explaining Performance Gaps in Medical Image Analysis Using Slice Discovery Methods. Olesen, Weng, Feragen, Petersen, 2024. Article
• Recommendations on compiling test datasets for evaluating artificial intelligence solutions in pathology. Homeyer et al. 2022. Article
• Responsible and Regulatory Conform Machine Learning for Medicine: A Survey of Challenges and Solutions. Petersen et al. 2022. Article
• 3D (c)GAN for Whole Body MR Synthesis. Mensing et al. 2022. Article
• Automatic Extraction of Rules for Generating Synthetic Patient Data From Real-World Population Data Using Glioblastoma as an Example. Appenzeller, Terzer, Homeyer, Redlich et al. 2025, Preprint

Our experts have profound expertise in processing large varieties of medical data, including medical images, claims data, registry data, clinical data (e.g. lab or parameters exported from HIS systems), medical reports, and sensor data.

We accompany the whole process from initial modelling, and setup of data collections, over data curation, augmentation or annotation, selecting proper AI models and assessment of model performance, to model inspection. Furthermore, we implement relevant technical standards and requirements to achieve legal and regulatory compliance.

Learn more about Services of Fraunhofer MEVIS.