AI-PROGNOSIS_D4.4 / The AI-PROGNOSIS digital health ecosystem (MVP)

Executive summary

D4.4 ‘The AI-PROGNOSIS Digital Health Ecosystem (MVP)’ delivers the Minimum Viable Product (MVP) version of the AI-PROGNOSIS’ three applications, i.e., mAI-Health, mAI-Care and mAI-Insights, which altogether form an ecosystem of digital health tools for supporting Parkinson’s Disease (PD) screening, disease management and clinical decision-making on medication selection and anticipated response. Building on the alpha versions delivered in D4.2 ‘The AI-PROGNOSIS Digital Health Ecosystem (Alpha version)’ and the technical specifications refined in D4.3 ‘Updated technical specifications, architecture and product backlog’, this deliverable describes the maturation of the ecosystem into functional MVPs ready for real-world proof-of-concept validation.

The transition from alpha to MVP was driven by updated user requirements and extensive co-creation activities with the patient panel and clinical partners, ensuring that both patient-facing and clinician-facing features are aligned with their needs and workflows. Most planned functionalities have now been implemented, including smartwatch data capturing, several self-reporting and self-logging assessments, the active cognitive and motor function tests, and initial feedback elements, such as digital biomarker graphs and clinician-facing reports on PD risk, progression, and medication response.

At the application level, the usage and components of the three applications are described, providing a clear overview of their functionality. The backend is underpinned by a robust, Kubernetes-based infrastructure that hosts data storage, analytics, and workflow orchestration. Apache Airflow coordinates the project’s data processing pipelines, which analyse wearable and clinical data to generate digital biomarkers and predictions of PD risk, progression, and medication response. A Large Language Model (LLM)-based explainability module has been further set up to generate clinician-facing lay summaries of predictions and associated explainability evidence. Monitoring and deployment processes follow GitOps principles to ensure stability, traceability, and scalability.

Overall, D4.4 demonstrates that the AI-PROGNOSIS digital health ecosystem has progressed from early prototypes to a coherent, integrated MVP towards supporting PD risk monitoring and patient management. The remaining product features and data processing pipeline refinements are underway and will be delivered with the final, refined MVPs as part of D4.5 ‘The AI-PROGNOSIS Digital Health Ecosystem (refined MVP)’.