VAMed

Project Description

VaMed addresses the growing need for efficient analysis of unstructured patient data in outpatient care. The project combines natural language processing and visual analytics to support decision-making in general practice. The goal is to reduce information overload and improve access to relevant clinical content through modular, interpretable visualizations. The system extracts medical concepts from free-text health documents, structures them, and maps them to task-specific visual components. These include timelines, entity cards, filters, and context-aware views. The modular architecture allows each visualization to operate independently but remain connected through a shared data model.

Data extraction is handled by transformer-based NLP models, including domain-specific variants such as BioBERT and LLaMA. Extracted entities are typed, scored, and matched to predefined visual mappings. The system processes documents such as referrals, discharge letters, and medication summaries. Outputs are rendered in a lightweight dashboard designed for low-friction integration into clinical routines.

VaMed is developed for outpatient use, where time and attention are limited. It avoids complex multi-view visualizations in favor of modular, focused components. The interface supports rapid orientation and task-driven interaction, targeting daily medical workflows rather than academic exploration. All data used during development is synthetic. Real records are not stored or processed. The system architecture supports full local deployment, role-based access control, and logging. It is aligned with GDPR and the EU Data Act, and designed for secure use within healthcare infrastructure. VaMed is currently in prototyping. A clinical user study with general practitioners is in preparation. Planned evaluations will include usability, information accuracy, decision speed, and integration into daily workflows.

Core Features

  • Extraction of diagnoses, medications, symptoms, and temporal markers from free-text

  • Modular visual interface for patient summaries

  • Scalable architecture with visual and data-layer separation

  • Support for German clinical language and multilingual expansion

  • Secure, offline processing with no external dependencies

Publications

  1. Medical Visual Analytics – An Interactive Approach for Analyzing Electronic Health Records, International Conference on Information Visualisation, 2024

  2. A Modular Visual Analytics Dashboard for Patient Health Data, Information Visualization special Issue Journal, 2025 (to appear)

  3. Medical Visual Analytics – Visual Decision-Support for Primary Care, International Conference on Information Visualisation,  2025 (to appear)

Our cooperation partner

Group Practice for General and Internal Medicine, Rodgau-Jügesheim, Germany

This project is funded by the Ministry of Science and the Arts (HMWK) within the framework of the academic mid-level program.