Periodic Reporting for period 1 - eBRAIN-Health (eBRAIN-Health - Actionable Multilevel Health Data)
Okres sprawozdawczy: 2022-07-01 do 2023-06-30
Biomedical research is currently hindered because technical infrastructure is missing to protect the privacy of personal data. Problematically, biomedical data cannot be easily anonymized or pseudonymized such that personally identifiable information are removed, and potential re-identification is excluded. Person-related information are the explicit target in biomedical research and for personalized therapy. Precisely that information that characterizes the biomedical state of a person can also be used to re-identify a person. Those features that create a better understanding of health and illness of a person are the ones that identify a person. Neuroimaging data, and especially data related to personalized brain network modelling, like structural connectomes, functional connectomes, brain surfaces, MRI images, fMRI time series, PET loadings, simulated brain activity, parameter estimation results, etc. can potentially identify the person from whom the data was recorded and therefore EU laws require proper protection of such data.
eBRAIN-Health main objective is to deliver a distributed research platform for modeling and simulating complex neurobiological phenomena of human brain function and dysfunction in a data protection compliant environment. The goal is to provide thousands of multilevel virtual brains - the eBRAIN-Health “twins" from patients and healthy human controls for research and innovation.Brain data from multiple sources will be pre-processed and annotated with a common data model - such that they all will relate to common spatial and temporal reference frameworks. The platform thus will offer a next generation clinical research infrastructure - compliant with the EU General Data Protection Regulations (GDPR) and creates an open yet protected space for groundbreaking digital health innovation by the research infrastructure communities comprising academia and the private sector. eBRAIN-Health utilizes neurodegenerative disorder data as a demonstration case. The project will integrate large, complex datasets and computational simulations from diverse academic medical research projects to advance our understanding of the underlying causes of these diseases, enabling identification of new potential treatment options. eBRAIN-Health systems can subsequently be applied as a benchmark/model for many other disease areas. The eBRAIN-Health infrastructure is designed to respect the sensitive nature of medical data, while making these data accessible for further research. Combined with access to high-performance computational power, the systems deployed in the project will allow generation of meaningful algorithms, models and simulations. The platform will be attractive not only to the academic research community but to the private sector. Start-ups and established drug developers can use the infrastructure for early digital/in-silico screening of new approaches, stress-testing them against models of real-world data. This can accelerate development of successful treatments or save resources through early identification of failures. From a disease management perspective, the predictive power provided by these models and simulations will help improve clinical care, e.g. by identifying which intervention is best suited for a particular individual.
eBRAIN-Health main objective is to deliver a distributed research platform for modeling and simulating complex neurobiological phenomena of human brain function and dysfunction in a data protection compliant environment. The goal is to provide thousands of multilevel virtual brains - the eBRAIN-Health “twins" from patients and healthy human controls for research and innovation.Brain data from multiple sources will be pre-processed and annotated with a common data model - such that they all will relate to common spatial and temporal reference frameworks. The platform thus will offer a next generation clinical research infrastructure - compliant with the EU General Data Protection Regulations (GDPR) and creates an open yet protected space for groundbreaking digital health innovation by the research infrastructure communities comprising academia and the private sector. eBRAIN-Health utilizes neurodegenerative disorder data as a demonstration case. The project will integrate large, complex datasets and computational simulations from diverse academic medical research projects to advance our understanding of the underlying causes of these diseases, enabling identification of new potential treatment options. eBRAIN-Health systems can subsequently be applied as a benchmark/model for many other disease areas. The eBRAIN-Health infrastructure is designed to respect the sensitive nature of medical data, while making these data accessible for further research. Combined with access to high-performance computational power, the systems deployed in the project will allow generation of meaningful algorithms, models and simulations. The platform will be attractive not only to the academic research community but to the private sector. Start-ups and established drug developers can use the infrastructure for early digital/in-silico screening of new approaches, stress-testing them against models of real-world data. This can accelerate development of successful treatments or save resources through early identification of failures. From a disease management perspective, the predictive power provided by these models and simulations will help improve clinical care, e.g. by identifying which intervention is best suited for a particular individual.
In project phase 1, eBRAIN-Health has worked towards solutions for the following key challenges:
1. Limited data findability, access and usability:
· Data Findability and Access: Systematic data-driven approaches are hampered because patient-level data are difficult to discover and access or, if the (raw) data are known to exist, they are entirely inaccessible.
· Data Models: Metadata models of different cohorts and/or data modalities use different terminologies and refer to differing definitions and concepts, limiting usability beyond individual projects.
2. Limited analytical skills and lack of linkage with knowledge bases beyond individual projects:
· Data Processing: Individual scientists are often limited in data modalities or features they can analyze from the available raw data due to limitations in expertise and in access to computational power.
· Fragmentation: Existing biological and medical knowledge and data are distributed (e.g. in articles, knowledge bases). Seeking information is time-consuming and retrieving relevant data is often not feasible.
· Theory: Complex mathematical models and simulations are required to infer biophysical principles that transcend single levels and can link multilevel data.
3. Lack of access to HPC and robust, scalable GDPR compliant infrastructure:
· Industry-grade service: Sustainable, scalable, reliable quality assured data, tools and infrastructure
· Data Protection: Technical and organizational means are necessary to provide data protection by design and by default. There is a need for digital Research Infrastructures that provide high-performance computing, storage and services for running workflows, as well as adequate technical security of services and storage.
4. Addressing ethical and social challenges linked to the development and use of digital twins for dementia, including:
· Representation, bias, consent, inclusivity of research and promotion of diversity
· Legal capacity, shared and supported decision-making, communication, trustworthiness, and trust.
1. Limited data findability, access and usability:
· Data Findability and Access: Systematic data-driven approaches are hampered because patient-level data are difficult to discover and access or, if the (raw) data are known to exist, they are entirely inaccessible.
· Data Models: Metadata models of different cohorts and/or data modalities use different terminologies and refer to differing definitions and concepts, limiting usability beyond individual projects.
2. Limited analytical skills and lack of linkage with knowledge bases beyond individual projects:
· Data Processing: Individual scientists are often limited in data modalities or features they can analyze from the available raw data due to limitations in expertise and in access to computational power.
· Fragmentation: Existing biological and medical knowledge and data are distributed (e.g. in articles, knowledge bases). Seeking information is time-consuming and retrieving relevant data is often not feasible.
· Theory: Complex mathematical models and simulations are required to infer biophysical principles that transcend single levels and can link multilevel data.
3. Lack of access to HPC and robust, scalable GDPR compliant infrastructure:
· Industry-grade service: Sustainable, scalable, reliable quality assured data, tools and infrastructure
· Data Protection: Technical and organizational means are necessary to provide data protection by design and by default. There is a need for digital Research Infrastructures that provide high-performance computing, storage and services for running workflows, as well as adequate technical security of services and storage.
4. Addressing ethical and social challenges linked to the development and use of digital twins for dementia, including:
· Representation, bias, consent, inclusivity of research and promotion of diversity
· Legal capacity, shared and supported decision-making, communication, trustworthiness, and trust.
eBRAIN-Health continues the development of EBRAINS (ebrains.eu) Service for Sensitive Data (SSD) – the Health Data Cloud (HDC). The HDC is an open-source data management and processing platform that has evolved from the EOSC project Virtual Brain Cloud (led by CHARITE) and produced a GDPR audited Virtual Research Environment that enables medical researchers to store, process and share data in compliance with the European Union (EU) General Data Protection Regulation (GDPR). The HDC addresses the present lack of digital research data infrastructures fulfilling the need for (a) data protection for sensitive data, (b) capability to process complex data such as radiologic imaging, (c) flexibility for creating own processing workflows, (d) access to high performance computing. The platform promotes FAIR data principles and reduces barriers to biomedical research and innovation. It offers a web portal with graphical and command-line interfaces, segregated data zones and organizational measures for lawful data onboarding, isolated computing environments where large teams can collaboratively process sensitive data privately, analytics workbench tools for processing, analyzing, and visualizing large datasets, automated ingestion of hospital data sources, project-specific data warehouses for structured storage and retrieval, graph databases to capture and query ontology-based metadata, provenance tracking, version control, and support for automated data extraction and indexing. The HDC is based on a modular and extendable state-of-the art cloud computing framework, a RESTful API, open developer meetings, hackathons, and comprehensive documentation for users, developers, and administrators. The HDC with its concerted technical and organizational measures can be adopted by other research communities and thus facilitates the development of a co-evolving interoperable platform ecosystem with an active research community. A prominent demonstrator of HDC’s functional scope is a recent study on simulations of 650 personalized human brain digital twins (Schirner, Deco, Ritter (2023) Nature Communications).