Fraunhofer Institute for Telecommunications,
Heinrich-Hertz-Institut, HHI
NewsTechnologies and SolutionsResearch GroupsProjectsPublicationsFinal ThesisPeople and Contact

AI-Solutions for Health

At the AI department of the Fraunhofer Heinrich-Hertz Institute (HHI), our researchers develop advanced machine learning methods to address a wide range of medical challenges. Our work spans the analysis of medical images, time series data, clinical texts, and biomedical sequences, as well as multimodal health data. 

Medical Imaging

Medical imaging enables the visualization of biological structures within the body using microscopes, X-ray, magnetic resonance imaging (MRI), dermatoscopes, or cameras. Our AI-driven methods facilitate image reconstruction and automated pathology detection, supporting clinicians with accurate and reliable insights. We prioritize explainable AI to ensure transparency and safety in medical applications.

Figure 1: Example of a histological slide. [Source: Wagner et al. (2025) Semantic modeling of cell damage prediction: a machine learning approach at human-level performance in dermatology.
Sci Rep 13. https://doi.org/10.1038/s41598-023-35370-7.
License: http://creativecommons.org/licenses/by/4.0/ ]

Medical Text

Large language models are transforming healthcare by enabling new applications, such as assessing guideline compliance and classifying clinical documents. We specialize in fine-tuning and evaluating and interpreting these models for hospital use and related applications, ensuring they meet the high standards required for medical environments.

Biomedical Sequences

The molecular language of life is encoded in sequences of nucleic acids and amino acids. Understanding the properties of proteins and peptides is essential for both fundamental research and medical applications, such as immunology. Since only a small fraction of proteins has been experimentally characterized, our research focuses on predicting their properties from primary sequences. We leverage recent advances in natural language processing to drive discovery in this field.

Figure 3: Explainable AI can help make the workings of protein language models more transparent in predicting protein functions. [Figure adapted from: Wenzel, Grüner, Strodthoff (2024). Insights into the inner workings of transformer models for protein function prediction, Bioinformatics, 40(3). https://doi.org/10.1093/bioinformatics/btae031. License: https://creativecommons.org/licenses/by/4.0/]

Projects

  • G-TERM-AI: Generative Testbed for Enhancing Device- and Protocol-Agnostic Reliability of Multimodal Medical AI.
  • AI/AR Microscope: Assistive Artificial Intelligence and Augmented Reality Tools for Microscopic Diagnostics in Dermatology and Microscopic Surgery.
  • TEF-Health: Testing and Experimentation Facility for Health AI and Robotics.
  • DAKI-FWS: Data- and AI-supported early warning system.
  • SyReal: Synthesizing realistic variations in data for reliable medical machine learning at scale.
  • iToBoS: Intelligent Total Body Scanner.
  • Patho234: Machine learning-driven multidimensional imaging analysis of reactive and neoplastic lymph nodes.
  • TraMeExCo: Transparent Medical Expert Companion.
  • FG-AI4H: ITU/WHO Focus Group on Artificial Intelligence for Health.

Publikationen

Medical Imaging

Medical Time Series

Biomedical Sequences

Various