Trustworthy AI development is a key part of the AI-PROGNOSIS approach to building digital health tools for Parkinson’s disease.

The updated AI-PROGNOSIS report presents the project’s Trustworthy AI Development and Evaluation Framework. It builds on the previous version and brings together updated practices, technical reporting, and guidance for the development and evaluation of AI tools within the project.

The framework covers key dimensions of trustworthy AI, including robustness, generalisation, explainability, transparency, reproducibility, fairness, privacy, and accountability. The updated version also introduces environmental sustainability as an additional dimension.

The report also considers relevant regulatory and reporting frameworks, including the EU AI Act, FDA guidance on AI-enabled medical devices, FUTURE-AI international consensus guidance, ISO/IEC standards, and the TRIPOD-AI checklist for transparent reporting of AI models in healthcare.

The updated framework further includes practices for explainable AI, LLM-based explanation pathways, and risk management across the AI lifecycle. It also consolidates reporting from model and application development partners on implementation progress and future plans.

This work supports the trustworthy development, evaluation, and future deployment of AI tools within the AI-PROGNOSIS digital health ecosystem.

Read the full report on Zenodo: https://zenodo.org/records/21102801