Work

D2.1. Data Management Plan

Year: 2025

Type: article

Abstract: The Data Management Plan (DMP) for the ChatMED project outlines a strategy for managing neurological data to develop Large Language Models (LLMs) for healthcare. Central to the DMP is the TriZ-Flow architecture, a three-zone data processing pipeline (Bronze, Silver, Gold Zones) ensuring data quality, standardization, and GDPR compliance. This decentralized approach minimizes privacy risks while enabling efficient data integration. The DMP emphasizes ethical data handling, including informed consent and data minimization, and adheres to FAIR principles (Findable, Accessible, Interoperable, Reusable) to ensure data is organized, secure, and reusable. It integrates multimodal data (e.g., neuroimaging, patient histories) to enhance AI-driven diagnostics in neurology. The DMP also serves as a training tool for researchers, particularly in Widening countries. It provides infrastructure for the Human Evaluation Framework (HEF), ensuring rigorous evaluation of LLMs in neurology tasks. Designed for scalability, the TriZ-Flow architecture extends beyond ChatMED, ensuring longterm impact. Robust data security measures, including Role-Based Access Control (RBAC) and Attribute-Based Access Control (ABAC), protect sensitive information. An Ethics Advisor oversees compliance with ethical guidelines.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.16900972