Building on this infrastructure, we design and validate artificial intelligence models that support clinical decision-making. In one of our projects, we developed a model for patients with myelodysplastic syndromes (MDS) that integrates longitudinal laboratory data with diagnostic information to predict individualized one-year mortality risk dynamically over time. Validations on external cohorts from our cooperation partners show the robustness and performance of the method while also retaining a high degree of interpretability. This example illustrates how the intelligent use of routine data can provide clinicians with valuable insights to guide patient management.

In addition to classical machine learning approaches, we increasingly employ large language models (LLMs) to, for example, extract structured information from unstructured clinical documents. This technology enables the efficient utilization of retrospective and external data sources, thereby expanding the available data basis for clinical and translational research.

Through these combined efforts, we contribute to the digital transformation of hematology and oncology by creating sustainable infrastructures, innovative analytical methods, and AI-driven applications that support precision medicine and improve patient outcomes.