FONDA - Grundlagen von Workflows für die Analyse großer naturwissenschaftlicher Daten

01.07.2024 bis 30.06.2028

Our subproject within the CRC1404 aims to improve the adaptability and reliability of machine learning (ML)-based biomedical image analysis workflows. It focuses on developing a Python-based domain-specific language (DSL) and modular tools to assess model suitability, adapt models to new datasets, implement transfer learning, and quantify uncertainty. Intuitive visualizations will support human evaluation of results, crucial for clinical applications. The approach will be validated in neuroimaging-based disease classification and cell segmentation in microscopy, enhancing workflow generalizability and reliability for biomedical research.