The EU project STRATIF-AI

Digital twins for stroke

A scalable platform for continuous stratification using AI and digital twins

State-of-the-art stratification today is based on machine-learning (ML) algorithms, trained on large cohort data. This has two main limitations: a) such ML-models cannot use all the variety of different data that is generated about a patient ,b) stratification is thus only done intermittently, implying out-dated and sub-optimal care decisions. To remedy this, we herein present a new concept and technology - continuous stratification, using our new STRATIF-AI platform.

World-unique digital twins

In continuous stratification, all data generated about a patient is cumulatively stored in a Personal Data Vault, controlled by the patient. These personal data continuously updates our world-unique digital twins. The unique potential with our twins comes from the hybrid architecture, combining mechanistic, multi-scale, and multi-organ models with ML and bioinformatics. This allows us to simulate patient-specific responses to changes in diet, exercise, and certain medications, and see changes on both an intracellular, organ, and whole-body level, ranging from seconds to years. We also combine semantic harmonization with federated learning to securely re-train the various sub-models, when new data become available in one of the cohort databases.

An interconnected and patient-centric healthcare system

In this project, we will for the first time use this cutting-edge technology to connect a series of apps that together covers an entire patient journey. Using 6 new clinical studies, involving 8 new partner hospitals, we will both refine and validate the models, and demonstrate how the same digital twin can follow a patient across different apps, covering all phases of stroke: from prevention, to acute treatment, and rehabilitation. Our scalable platform for continuous stratification forms the foundation for a new interconnected and patient-centric healthcare system.

Work Packages

WP1 - Semantic interoperability and federated learning platform

Lead: Tree Technology SA (ES)
Duration: M1 - M48

The aim of this package is twofold: (1) to design and develop original training procedures for a wide variety of algorithms under the different privacy operation models and (2) to implement the semantic harmonization layer to allow the meaningful re-use of data from hospitals and research centers for building personalized health computational models. A FAIR Data Catalogue will contribute to data interoperability by facilitating data discoverability and exploration for data scientists and researchers. This WP also includes the creation of a data management plan, a specification of connections between the different data sources, and data pre-processing (including pseudo-anonymization), normalization and alignment of horizontal and vertical distributed datasets. Finally, this platform includes the development of a federated learning platform, which is used in WP2, for training the ML models.

WP2 - STRATIF-AI platform development and hybrid modelling

Lead: Software Imagination & Vision SRL (RO)
Duration: M1 - M42

This WP is devoted to the construction of the STRATIF-AI technical platform. The platform consists of several existing apps and technologies, which are used across the different phases of stroke. These apps are now connected into a single platform, by having them all talk to a shared backend, which contains the digital twin models and a personal data vault. In this WP, we will specify the technical details of all these components, and their interactions, develop the components, and merge them into the overall platform.

WP3 - Ethical and trustworthy design of STRATIF-AI platform

Lead: Charité - Universitätsmedizin Berlin (DE)
Duration: M1- M48

This WP corresponds to the development, testing, and validation of the ethical aspects of the platform. A plan for how these ethical and trustworthiness aspects will be tested will be developed, and then used throughout the project. Apart from this, WP3 also deals with submission of ethical approval documents, and with regulatory agencies, which ultimately will approve the legal and security aspects of the various components in the interconnected platform.

WP4 - Prevention

Lead: Region Västerbotten (SE)
Duration: M13 - M48

The overall objective with this WP is to develop, validate and evaluate the models and technologies developed for continuous stratification during prevention of stroke. More specifically, we will: a) validate the detailed physiological models, using a new type of multi-modal longitudinal data collected within the project, b) fine-tune and mature the prototype technology, to be suited for both primary and secondary prevention in several countries, c) do a randomized control trial to see differences in preventive care with and without the new technology, d) quantify these differences.

WP5 - Acute treatment and monitoring

Lead: Centre Hospitalier Regional et Universitaire De Brest (FR)
Duration: M13 - M42

The overall objective with this WP is to develop, validate and evaluate the models and technologies developed for continuous stratification during the acute treatment and monitoring of stroke. More specifically, we will: a) improve models that help guide the primary intervention, b) develop new hybrid physiologically based models for the postacute monitoring of patients, c) gather data for all of these models, by complementing our large clinical databases at all hospitals, with new collections of data from other registries or from storage of data from new patients, d) co-design and refine new and existing prototype technology, and clinical workflows, together with end-users, c) do a new type of longitudinal time-resolved study, starting in the acute phase, and continuing through rehabilitation and treatment of the chronic condition (shared with WP6).

WP6 - Rehabilitation and follow-up of chronic condition

Lead: Sheffield Teaching Hospitals, NHS Foundation Trust (UK)
Duration: M13 - M48

The overall objective with this WP is to develop, validate and evaluate the models and technologies developed for continuous stratification during rehabilitation and follow-up of the chronic condition. More specifically, we will: a) improve models that help with prediction of cognitive function after rehab is over, by using all of the continuously available data, b) adopt hybrid physiologically based models for exercise and movements, to be applicable to physical rehab, c) gather the time-resolved data for all of these models, by adding patient data to our existing clinical databases, d) co-design and refine of new and existing prototype technology, and clinical workflows, together with end-users, e) do a pilot study of a smaller group of patients, which bring data from the acute phase, into rehab and follow-up of chronic condition.

WP7 - Coordination, dissemination, implementation, and exploitation

Lead: Linköping University (SE)
Duration: M1 - M48

The overall objective with this WP is to manage and monitor the implementation, progress and achievements of the project; to interact with the European Commission and to coordinate the different activities, the exchange of information and communication, dissemination, exploitation and use activities; develop implementation strategies in partnership with key stakeholders. The overall management covers non-scientific, knowledge and innovation, ethical (lifted out to WP3), regulatory, financial, administrative, and legal aspects. This comprises a) Continuous monitoring of the project’s progress and timely initiation of corrective actions; b) Scheduling and organising project meetings, coordinating its organisation and execution, and/or participation of the project in various external or self-organised events; c) Defining and implementing the communication procedures to be followed within the project and with external agents. Electronic services for communications will be defined, d) Establishing communication and reporting channels to the European Commission; e) Conduction of continuous quality assurance activities for the operation of the project and the production of its scientific and technical results within its lifespan.

Funded by Horizon Europe

STRATIF-AI has received funding over four years by Horizon Europe, the EU’s key funding programme for research and innovation, under grant agreement No 101080875.