COMFORTage Virtual AI-based Healthcare Platform Architecture

The Virtual AI-based Healthcare Platform (VHP) is the cornerstone of the COMFORTage project. Its purpose will be to enhance and support the management of dementia and frailty through integrated and personalized solutions aimed at patients, clinicians and caregivers. A number of components collectively form the backbone of this VHP, working together to offer a secure and efficient healthcare solution.

System architecture design of the AI-based Virtual Health Platform system

The design of the platform’s system architecture is the result of activities carried out as part of Task T2.5, which brought together technical partners responsible for the development of the platform’s components to contribute their knowledge and expertise, ensuring a robust and comprehensive design.

Development Process

The process to design the VHP system architecture began in April 2024 (month 4 of the project) with partners providing input regarding the interfaces of their components to ensure the seamless communication and interoperability among components.Following this, the desired functions and properties of each component were carefully defined so as to describe the features and attributes important for the platform to possess or demonstrate for a functional and efficient system. Finally, the partners provided detailed specifications based on these functions and properties with respect to open interfaces, communication patterns, information flows and design principles. These specifications were defined to align with common standards, protocols and technologies, as well as the overarching goals of the Virtual AI-based Healthcare Platform.

Consolidation and Final Architecture

Once all the input was collected, it was consolidated into a unified system architecture in the form of a component interaction diagram. This consolidation process was essential in order to harmonize the diverse contributions and ensure that the resulting architecture not only fulfils the technical requirements but also aligns with the project’s vision of delivering a cutting-edge healthcare solution. A full report of these activities can be found in Deliverable D2.8 – “Reference Architecture and Integration of VHP Platform I”, delivered in December 2024 (month 12 of the project).

Key Components of the VHP System Architecture

The VHP system architecture is built around several key components, each playing a vital role in achieving the platform’s objectives:

• Data Ingestion and Harmonization

A number of components will be responsible for ensuring that data from diverse sources, including wearables, medical devices and electronic health records, are properly collected and processed by applying cleaning, normalization, standardization and transformation functions. By harmonizing these data into holistic healthcare records, the platform will enable seamless integration and usability across its ecosystem.

• Data Lake Management

Several centralized data repositories will form the backbone of the VHP data lake, enabling efficient storage, retrieval and analysis of vast amounts of healthcare data. The data lakes will facilitate advanced analytics and will support the development of personalized care strategies. A Blockchain Information Exchange component will also be available to provide strong cyber-security, regulate data access and to facilitate the secure and trustful exchange of data.

• Patient Treatment and Monitoring Tools

These tools will provide real-time insights into patient health, allowing for proactive interventions and continuous monitoring. They will be designed to improve patient outcomes through tailored treatment plans. Specifically, the Clinical Decision Support Suite (CDSS) will serve as the main entry point and frontend application of the platform, providing healthcare professionals access to the various integrated care features such as query building, data analysis and visualizations. The personalized Patient Digital Twins (PDTs) will provide clinicians and medical experts with personalized virtual models built from biological and physiological data of dementia patients, with which they will be able to view medical histories and generate individualized care recommendations.

• Integrated Care Model Library (ICML) and Explainable AI (XAI)

The ICML will offer predictive models and data analytics, while XAI will ensure that these models are transparent and interpretable. This combination will empower clinicians to make informed decisions concerning patient care. The components will employ several machine learning, deep learning and reinforcement learning techniques and algorithms so that users will be able to carry out predictive analyses regarding dementia and frailty prognosis, prevention and delay of disease progression at a personalized level.

• Training and Educational Platform

Recognizing the importance of citizen engagement, the Virtual AI-based Healthcare Platform will include a training and educational platform to enhance digital literacy and to raise awareness. User-centric educational guidelines and resources, as well as training materials and solutions will be offered through a Training and Educational Toolkit (TET), alongside a variety of serious games intended for patients focusing on brain training activities to improve cognitive skills on a daily basis. The TET and serious game applications will be hosted in a secure and personalized Training and Educational Marketplace (TEM). The TEM will be powered by a Trainings Recommender that will provide patient users with personalized suggestions regarding the content and elements of the serious game to improve their learning experience.

• People-Centric Assistive Applications

These applications will be designed to enhance user experience and support personalized healthcare delivery. Examples include virtual assistants and opt-in tools tailored to individual needs. Their goal will be to provide recommendations via a range of edge devices, such as smartphones, tablets and wearables, targeting various prevention and intervention measures based on the results of the models of the PDTs and ICML components. These recommendations will be in the form of reminders and alerts to support patients in their everyday activities.

• Authentication and Authorization Mechanisms

Robust security measures are integral to the VHP, ensuring that only authorized users have access to sensitive healthcare data and functionalities. For this reason, the platform will incorporate several mechanisms to securely govern access to the platform itself and its resources (data, models, applications and other assets).

The VHP system architecture demonstrates the power of collaboration and innovation in research for addressing complex healthcare challenges. By bringing together diverse expertise and leveraging cutting-edge technologies, the COMFORTage project has designed the architecture of an AI-intensive platform that will prioritize user-centric solutions. This unified approach ensures that the VHP is well-equipped to support the prevention, monitoring and management of dementia and frailty, ultimately enhancing the quality of life for individuals across Europe.

**Article written by Constantinos Stylianou, Andreas Andreou, Efthyvoulos Kyriacou, Andreas Christoforou, from Cyprus University of Technology (CUT), a key partner in the COMFORTage project.