Digital Biomarkers

Digital biomarkers are quantifiable patient parameters that enable computer-aided targeted diagnostics and personalized medicine.

Our research focusses on developing AI tools to analyze pathology images and to discover novel biomarkers. It spans the entire development cycle, from exploratory scientific research through the development of usable prototypical software tools to translation into marketable in vitro diagnostics (IVDs).

The resulting software solutions support disease diagnostics, reduce the workload of pathologists, enable more effective treatment, and ultimately lower healthcare costs.

Meaningful Digital Biomarkers

Histopathology images contain a high density of morphological and molecular information in its spatial context. Developing algorithms to quantify the complex interplay of different molecular mechanisms allows exploring clinically meaningful biomarkers in a data-driven way. Read more

Foundation Models—Learning from Small Datasets

Images in histology are highly variable, such that large datasets are usually needed to cover this variability when training algorithms. Collecting such datasets is challenging, in particular for small patient collectives. Foundation models aim to solve this dilemma by integrating multi-center knowledge for AI in personalized medicine. This allows building robust specialized models even if only small training datasets are available. Read more

Hemato-Oncology

Bone marrow smears contain a plethora of different cell types that need to be assessed together with a patient’s comprehensive data. The development of AI-driven algorithms for the analysis of blood and bone marrow cells supports differential cell counts algorithmically to diagnose hemato-onlological diseases. Read more

Image Registration

Comprehensive assessment in histopathology often requires different stains in adjacent tissue slides. Image registration fuses this this multi-stain information via image analysis. It allows the integration of different antibodies in the development of new image-based biomarkers and also serves as a tool to accelerate clinical routine diagnosis. Read more

Development of Software IVDs

Research prototypes need to mature in various ways to meet the quality, efficiency, and regulatory requirements of in-vitro diagnostic medical devices. Drawing on many years of experience in developing software solutions for clinical practice, this expertise enables the integration of technically validated algorithms into interoperable tools for routine use. Read more

Data Acquisition and Curation

Fraunhofer MEVIS has expertise in selecting representative data collections, which is crucial for unbiased AI development and evaluation. In clinical use, different systems need to work together. We develop interfaces adhering to standards like DICOM and ensure the interoperability of systems and datasets.

We partner with clinical, industry, and academic organizations to deliver innovative, data-driven solutions for biomedical research and diagnostics. Our experienced interdisciplinary team combines scientific, clinical, and technical expertise to support joint research collaborations and contract R&D projects—from early concept development to validated applications.

We offer cooperations in the following areas:

Digital Biomarker R&D – Data-driven identification of novel biomarkers based on images and other clinical data for clinical and pharmaceutical research studies.

Image & Data Analysis – Algorithms for robust qualitative and quantitative analysis of microscopic images and clinically relevant data.

Software IVD Development – Practical software solutions and translation into marketable in vitro diagnostics.

Image Registration – Ready-to-use, quality-assured, and award-winning image registration software components to align and combine serial sections. Read more

Publications

• Data-Driven Discovery of Immune Contexture Biomarkers. Schwen et al. 2018. Article
• Artificial Intelligence in Pathology: From Prototype to Product. Homeyer et al. 2022. Article
• Recommendations on compiling test datasets for evaluating artificial intelligence solutions in pathology. Homeyer et al. 2022. Article
• The NCI Imaging Data Commons as a platform for reproducible research in computational pathology. Schacherer et al. 2023. Article
• Overcoming data scarcity in biomedical imaging with a foundational multi-task model. Schaefer et al. 2024. Article

Challenges

• 1^st place in the UNICORN MICCAI 2025 Lighthouse Challenge for the best overall medical foundation model as well as in the sub-categories pathology and radiology.
• 1^st place in the ANHIR challenge on histological image registration. 
• 2^nd place in the LEOPARD challenge for prostate cancer risk prediction.