TIER 1
Low compliance with the Data Federation Framework
- Public metadata catalogue search
The goal of EUropean Federation for CAncer IMages (EUCAIM) is to build Cancer Image Europe, a pan-European digital hybrid federated infrastructure of cancer-related radiological and nuclear medicine images and related information, ready to be used to develop Artificial Intelligence (AI) tools for Precision Medicine. EUCAIM addresses the fragmentation of existing cancer image repositories by building a distributed, federated Atlas of Cancer Images, which will include both common and rare types of cancer. Its mission is to streamline access to de-identified, high-quality real-world data, and foster collaboration among clinicians, researchers, and AI innovators. The project aims to accelerate the development, training, and benchmarking of AI-powered tools for cancer management, empowering healthcare professionals with data-driven tools for diagnosis and treatment, and ultimately enhancing patient outcomes.
The EUCAIM project is now launching an open call for new beneficiaries to join the consortium.
The objective of the EUCAIM open call is to:
To reach the above objectives, the new beneficiaries will implement data incorporation, scientific and/or clinical use cases.
A use case refers to a specific scenario or situation in which the Cancer Image Europe platform is used to address a real-life scientific or clinical question where access and re-use of large amounts of cancer data can improve patient outcomes1. The integration of use cases into the EUCAIM project will outline the technical, ethical and legal steps involved in the process, and will play a crucial role in defining requirements for the implementation of additional cancer image datasets, tools and AI algorithms to improve clinical outcomes.
1 Example of use case: Identification of imaging biomarkers for early detection of breast cancer.
Scenario: The use case focuses on leveraging the Cancer Image Europe platform to enhance the early detection of breast cancer through the identification of imaging biomarkers. The objective is to develop a robust and efficient AI algorithm to help radiologists in identifying potential malignancies at an early stage.
Steps: 1) Data collection and integration; 2) Training the AI model; 3) Validation of the AI model; 4) Integration in clinical workflows
Expected outcomes: 1) Improved early detection of breast cancer through the identification of relevant imaging biomarkers; 2) Enhanced efficiency in radiology workflows, reducing the time required for manual review; 3) Increased accuracy in distinguishing between benign and malignant lesions, reducing false positives and unnecessary interventions; 4) Empowerment of healthcare professionals with a valuable decision support tool for more informed clinical decisions.
Who could apply: 1) A data holder, e.g. a hospital with a repository of breast imaging data (mammography, MR and/or ultrasound images) committed to advancing early cancer detection, intending to incorporate this imaging data into the Cancer Image Europe Infrastructure to contribute to the development of a robust breast cancer detection model; 2) AI developers, e.g. a MedTech SME specialised in the development of AI algorithms for healthcare applications aiming to leverage the Cancer Image Europe infrastructure for the development, training, and validation of advanced AI algorithms focused on breast cancer diagnostics.
The following use cases are accepted for this call:
You can find an always-up-to-date overview of available data in the Cancer Image Europe Platfrom’s catalogue. You can reach the catalogue through the dashboard on the frontpage, or via direct link.
Applicants are expected to implement their proposed work within the duration of the EUCAIM project, which is scheduled to conclude in December 2026. The prospective implementation duration for applicants is therefore expected to be between 18-24 months.
The total maximum budget available for this call is 3,852,000 EUR (1,926,000 EUR EC contribution). The indicative proposed budget is 214,000 EUR per beneficiary. This includes 7% indirect cost and 50% co-funding rate, i.e. 107,000 EUR EC contribution per beneficiary. However, this does not preclude the submission/selection of proposals requesting other amounts. It is envisaged to invite approximately 18 new beneficiaries to join the EUCAIM consortium. The new beneficiaries will receive funding under the same co-funding conditions as consortium partners (i.e., 50% of their costs must be co-funded by the prospective beneficiaries), after formal inclusion in the EUCAIM consortium through an amendment to the Grant Agreement.
Applicants must be legal entities and may belong to one or both of the following groups:
Data holders: organisations that supply the data.
A data holder represents any legal person or entity, who has the right, obligation, or capability to make certain data available for research purposes (i.e. hospitals or healthcare providers, academic institutes, research institutes, private companies, small and medium-sized enterprises (SMEs), data repositories, regional biobanks, clinical centres, cancer screening programmes, public entities, pharmaceutical companies and data altruism initiatives). Data holders will contribute with data either by; (a) becoming a federated node or (b) transferring anonymized data directly to the Central Repository.
Data users wishing to develop, train, benchmark and/or validate AI algorithms using the curated data in the Cancer Image Europe platform.
Data users may include:
Researchers with a clinical scientific question from the real world that can be addressed through the use of the Cancer Image Europe platform. In this call, applications focused on the development, training, and validation of AI algorithms to answer a specific clinical need (use case) will be accepted (for example, for the identification of imaging biomarkers or the assessment and prediction of response to treatment). Details on requirements for access to and use of the data should be described in the application (local use or federated processing).
Private companies and small and medium-sized enterprises (SMEs) developing AI tools for improving diagnosis and prevention and wishing to train and/or validate their tools on external datasets.
Applications must be submitted online using the EUCAIM Open Call Application Form.
Deadline for submission is 10 June 2024.
During the submission, applicants must also upload the following supporting documentation:
Other documentation is requested but can be provided later if not available yet at the time of submission:
The applicants must state in the application form if some documentation is not available yet.
Applicants can contact [email protected] for further details and assistance during the application process.
Applications will undergo a first eligibility check based on the eligibility criteria defined in Annex 1. Eligible applications will then be evaluated by the EUCAIM Access Committee, with support from the EUCAIM legal, ethics and technical experts, based on compliance with the ethical, legal, scientific, and technical requirements described in Annex 1. The highest ranked applications will be shortlisted and recommended for accession to the EUCAIM Grant Agreement.
For the shortlisted applicants, the European Commission will perform formal eligibility checks (including legal, financial and ownership control checks, as required). Only applicants who fulfil the formal eligibility criteria outlined in the call document of the original call DIGITAL-2022-CLOUD-AI-02 (see section 1 (I) of Annex 1) can accede to the EUCAIM Grant Agreement. The final decision on the formal eligibility lies with the European Commission.
Applicants from widening countries are encouraged to apply. Widening countries are Bulgaria, Croatia, Cyprus, Czechia, Estonia, Greece, Hungary, Latvia, Lithuania, Malta, Poland, Portugal, Romania, Slovakia, Slovenia and countries associated to the Digital Europe Programme
2 Applicants will be requested to provide an Ownership and Control Declaration. The Ownership and Controler Declaration can be found here.
At the moment of application, all applicants must ensure that:
Before final approval of applications, applicants will be required to provide evidence of the above.
Successful applicants are committed to provide the related documentation (in English or including an English translation), to sign all the relevant agreements, and to be available to collaborate in ethical verification activities providing support to data access requesters, if needed. The compliance with this requirement is compulsory.
Applicants requiring access to EUCAIM data for the development of their use cases must also:
Furthermore, the AI tools must comply with current applicable European and national legislation (Regulation (EU) 2016/679 GDPR, national privacy legislations).
The processing of data, whether anonymized or pseudonymized, may require executing different legal agreements, including for example a data transfer agreement and a data processing agreement. The applicants must commit to provide any additional information and documentation upon request, and to sign all the relevant agreements.
EUCAIM will organise a training session on the ethical and legal requirements that must be fulfilled when becoming part of the Consortium for data sharing or re-use, right after the submission deadline. All the applicants will be duly informed about the details and are strongly encouraged to participate in this training activity.
Applicants will be evaluated similarly to prospective consortium members during the Grant Agreement evaluation stage. Given the scope and expected contribution of new applicants, the scientific quality of the applications will be evaluated based on the following criteria:
Relevance (alignment with the objectives and activities as described in this call document)
1. Scientific and clinical relevance.
A scientific/clinical use case should address a relevant clinical question and have an impact on improving patient outcomes. The use case is considered scientifically relevant, if it can help expand the knowledge base, advance the understanding of a certain subject and/or provide interdisciplinary insight.
A use case on technical development of AI tools should have a clear definition of the end product, including the intended use of the AI technology and potential areas of application in the clinics.
For a data incorporation use case, the scientific and clinical relevance resides in the potential use of the datasets to address relevant scientific and/or clinical questions.
2. Coverage of topics/areas of interest.
3. Soundness of the concept, and credibility of the proposed methodology
4. Clarity and pertinence of the objectives
Implementation (maturity of the proposal/use case)
5. Soundness of the implementation plan and efficient use of resources
6. Capacity of the applicants to conduct the proposed work (use case implementation)
7. For scientific/clinical use cases only: Soundness of the concept, and credibility of the proposed methodology; clarity and pertinence of the objectives.
8. For data holders only: types of data (imaging only or imaging data integrated with other clinical data); number of datasets; single or multi centre collection.
9. For data holders: Level of compliance with the EUCAIM Data Federation Framework (DFF, see annex 2). For AI use cases: maturity level. The use case can refer to the development, training/benchmarking, or validation of AI algorithms.
Impact (extent to which the expected outcomes and deliverables referred to in this call will be achieved)
10. Impact and innovation: any substantial impacts that would enhance innovation capacity, address issues related to cancer diagnosis, treatment, and prevention, bring important benefits for patients or healthcare systems, create new market opportunities and/or strengthen competitiveness and growth of companies.
Accepted data types
Types and formats of image data:
Other clinical data:
Other relevant information might be from the following categories: demographics, lab results, symptoms, history of the patient that relates to the cancer diagnosis, tumour markers, diagnostic procedures, radiology report information (e.g. PI-RADS, BI-RADS, etc.), histopathology report (e.g. TNM stage), treatments (e.g. surgeries, medications, treatment response, follow-ups, metastatic episodes), side effects.
Functional and technical requirements for data holders
For inclusion in the Cancer Image Europe platform, datasets should meet the following requirements:
To allow the smoothest possible onboarding process for data holders, full compliance with the Data Federation Framework (DFF) requirements defined by EUCAIM is not mandatory, and for the successful applications EUCAIM will incentivize data curation and facilitate data transformation, if necessary, to align with the DFF. Nevertheless, a higher level of compliance with the DFF (i.e. higher Tier, as defined in annex 2) will be judged positively for the final scoring of applications.
The successful applicants will be required to fulfil some technical requirements. Data holders who agree to establish their own data node will need to procure or own the necessary infrastructure, following local procurement procedures to obtain management and technical resources. Additionally, acquiring and installing dedicated hardware is necessary to meet specific processing demands, categorised into participation tiers described in annex 2. Data storage must align with EUCAIM’s specifications, either within or attached to the federated node(s). Furthermore, operating system installation and setup compatible with EUCAIM’s software stack, along with required services, are mandatory. Network configuration, including a wired internet connection and security protocol integration, is imperative for each federated node. Additionally, adherence to security measures, integration with traceability mechanisms, and provision of digital access to authorised personnel are essential. Maintaining data redundancy, ensuring minimum uptime, signing Service Level Agreements, securing physical infrastructure, and hosting nodes in restricted access zones are important requirements. Organisations unable to accommodate a node can use EUCAIM’s Central Storage for GDPR-compliant data sharing. Technical requirements include access to a dedicated machine, HTTPS network configuration, technical support, and compliance with deployment guidelines.
Technical Requirements
Minimum Documentation Requirements and Benchmarking Information
To promote transparency and facilitate the AI use case evaluation, AI tools developers are required to provide the following documentation and benchmarking information:
Furthermore, AI Tool developers must provide the following information about their tools:
Compliance and Certification Documents are not mandatory for the application but will be required from successful applicants.
References
4 For example, the MR procedure occurred 25 days after the tumour incidence (if it occurred after a biopsy for instance) or -10 days if the MR preceded the finding (the MR examination occurred before the actual diagnosis). This is mandatory only if the diagnosis date and imaging procedure occurrence date have been modified due to de-identification procedures. The relative days between confirmed diagnosis and examinations must be available.
The following section illustrates EUCAIM’s tier structure, i.e. the different levels of compliance for data to be made available through the platform.
Data will be accepted by the Federation, without compliance requirements concerning the source repository (mainly linked to an existing research project in the European framework). Data quality specifications established by the clinical centre of origin (in case of a clinical environment) will be checked against EUCAIM minimum data quality standards. Tier 1 data can either be shared from a federated node or transferred to the central repository. The functionalities offered by the EUCAIM platform will be limited accordingly: only publication and visualisation of the dataset in the public metadata catalogue will be possible, allowing basic centralised filtering. The data does not have to comply with the common data formats (EUCAIM’s Hyper-ontology). Neither federated/distributed processing capabilities nor a homogeneous framework for research will be available. The datasets in the public catalogue will be listed and made accessible (under the defined data request process). Data users will be warned that the data use is under these conditions.
Summary Tier 1:
Compliance with EUCAIM’s Federated Query service requires a stronger involvement of the DHs. It allows for improved visibility and usability of the data. Federated queries require provision of (meta)data according to EUCAIM’s Hyper-ontology, and/or operating a local “mediator” service to execute the federated queries and report back the aggregated results. Datasets adhere to the standardised EUCAIM common data model (CDM), making it easier for researchers to use multiple datasets from Tier 2 in their projects compared to those from Tier 1.
Summary Tier 2:
Full data compliance entails alignment with a wide set of requirements including data harmonisation, annotation, and quality assessment. Tier 3 compliance should be the ultimate goal for DHs and RCs to achieve best possible usability and impact of their datasets within the Federation. It goes beyond the Federated Query capabilities of Tier 2 and enables federated processing, including Machine Learning (ML) and other advanced data processing techniques.
Summary Tier 3:
The following tables illustrates availability and findability of data according to tiers:
EUCAIM is looking for your feedback! We have recently published a Stakeholder Survey in order to reach out to potential end-users and stakeholders. We believe that your insights could significantly contribute to understanding the expectations of potential users and identifying the essential aspects that stakeholders find crucial for future engagement and collaboration with the platform.
Therefore, we would like to invite you to participate in the Stakeholder Survey about the Cancer Image Europe platform.
Completing the survey will take approximately 10 minutes. Your participation is crucial to the success of this project, and we deeply appreciate your expertise in shaping the future of cancer imaging and treatment.