WP1: Ethics Requirements
This work package sets out the ethics requirements definded by the European Commision that the project must comply with.
WP2: Legal and Ethical Aspects
WP2 superivses all legal and ethical aspects of the repository’s design and intended operation modes. It is led by the project’s Data Protection Officer (UV). The specific aims are:
- To closely coordinate with the data protection management task in WP11, to assess, implement and audit privacy protocols at each participant entity.
- To construct the legal operating model of CHAIMELEON repository.
- To verify its correct implementation at the different stages of the repository functional prototype.
- To closely follow up the evolution of the regulatory frameworks on the research use of health data in the EU nations.
- To work proactively with European agents and national agents, in order to contribute to defining paths within the existing European regulatory framework of the GDPR, to facilitate more open transborder reuse of health imaging data across European nations.
WP3: Design of CHAIMELEON Repository
WP3 is concerened with the design of a EU-wide interoperable repository of health images and related clinical and molecular data. The WP will work in close collaboration with WP2. The repository will be set up as a distributed infrastructure with high flexibility, scalability and efficiency to deal with a flexible workload. It is led by UM. The specific aims are:
- To ensure the implementation of the necessary security, confidentiality and integrity means to fulfil the requirements of the GDPR, according to the findings in WP2.
- To support the repository with a set of distributed services that will provide high availability and efficiency to deal with a flexible workload.
WP4: Set-up of CHAIMELEON Repository
WP4 is concerned with the set-up of the repository, supporting access to health imaging (and related) datasets and online processing tools/ pipelines. It is led by UPV. The specific aims are:
- To provide a data repository infrastructure supporting both access and online processing.
- To ensure the long-term preservation of the data repository.
- To provide high-availability services, efficient access, long-term data preservation and security.
- To be compatible with existing distributed infrastructures for research funded by the European Commission and regional initiatives.
- To support repeatability of the experiments fostering the development of reproducible science.
WP5: CHAIMELEON Population with Data
In WP5, the CHAIMELEON repository will be populated with quality-checked imaging data used for precise diganosis, prognosis and follow-up of lung. breast colorectal, and prostate cancer. The data will be provided by project parterns and external collaborators. The WP inlcudes the implementation of the required infrastructure at each data provider center for data pseudonymisation, incorporation and curation, according to common protocols established for the repository. The WP is led by CERF. Its specific objectives are:
- To populate the repository with new cases (complete imaging studies including clinical data) corresponding to patients diagnosed in the participant hospitals for the 4 target types of cancer, collected through the routine delivery of health care, with no enrolment conditions (Real World Data).
- To implement the tools for imaging data and clinical reports de-identification/pseudonymisation,
- To expand the Repository to reach a wider EU coverage by the implication of external collaborators, implementing interfaces with their hospitals data warehouses.
- To enable accessibility of curated data for estimated 38.000 cases provided by Partners and 1.200 cases by Collaborators (estimated 30% for lung, 30% colorectal,15% breast, 25% prostate cancers).
- To work in close coordination with WP1, ensuring that all the protocols, methodologies and tools delivered are in compliance with regulatory requirements.
WP6: Approaches for Harmonisation of Multicentric Datasets
WP6 develops approaches and tools for enhancing the reproducibility of radiomics features for cross-vendor/cross-institution image datasets. The most effective models will be implemented as an online pipeline accessible in the repository for harmonisation of multicentre datasets. It is led by IMPERIAL. The specifc objectives are:
- To explore different approaches for enhancing the reproducibility of radiomics features for cross-vendor/cross-institution image datasets.
- To prioritise approaches which move away from optimising images for their analysis by a radiologist towards optimising images for their computational analysis (extraction of QIBs using AI).
- To identify novel acquisition methods, such as fast, quantitative multiparametric MR mapping for facilitating the delivery of standardized QIBs.
- To explore disruptive approaches based on the generation of synthetic images with a common QIB conserving format, focused on stability of quantitative features instead of perceived image quality.
- To undertake a comparative assessment of the effectiveness of the different approaches for enhancing reproducibility of extracted QIBs.
WP7: Approaches for Integrity and Interpretability of AI solutions
WP7 aims to develop and implement approaches and tools for assisting the development of AI-based solutions, including data privacy techniques to prevent possible reverse engineering of a patient’s identity, tools for securing the authenticity and integrity of synthetic health images, methodologies for facilitating the interpretability of AI models, among others. It is led by BGU. The specific objectives are:
- To deliver tools for detecting changes introduced by the AI algorithms and to secure the authenticity and integrity of the synthetic images while in transit (over a network) and at rest (in storage).
- To deliver an analytical engine assist in the selection of AI training/testing datasets.
- To deliver methodologies and tools for moving towards explainable AI for users to better infer reasons behind the models’ outputs and/or to have more understanding on the criteria for consider such outputs as reliable.
- To explore new approaches based on federated/distributed learning to facilitate data privacy compliance multicentre data sharing and analysis, particularly important in AI developments.
WP8: Internal and External Validation of CHAIMELEON Repository
WP8 will undertake the internal and external validation of the CHAIMELEON repository and the tools for the assessment of their usefulness. The WP is led by GE Healthcare. The specific objectives are:
- To perform the internal validation of the repository carried out by AI developers in the consortium as the assessment of how useful the CHAIMELEON repository is for (i) facilitating experimentation of AI solutions and (ii) contributing to better training and testing of the AI models.
- To facilitate the external validation of the repository by other world-class European AI developers for health applications articulated as CHAIMELEON Open Challenges to the community consisting of experimentation with AI tools for cancer management using CHAIMELEON repository resources for a period of approx. 6 months. The challenges will be awarded to the higher evaluated tools for their relevance in addressing a specified clinical end point of high relevance in the management of each of our targeted types of cancer cancers.
WP9: In-silico External Validation of AI Solutions for Cancer Management
WP9 is concerned with the early validation of how the proposed AI-based solutions can contribute in addressing a specified clinical end point of high relevance in the management of each of our targeted types of cancer cancers. Observational (non-interventional) studies are designed and executed under the leadership of CERF (for lung cancer), ULS (for colorectal cancer), PSD (for breast cancer) and HULAFE (for prostate cancer). This WP also agglutinates the contribution towards international standards for AI-based solutions in health imaging. It is led by HULAFE. The specific objectives are:
- To perform in-silico external validation of our AI tools with a dataset cohort of patients different from the one used for their training and internal validation in WP8, for the 4 target types of cancer, collected through the routine delivery of health care with no enrolment conditions (Real World Data).
- To undertake the in-silico external validation as non-interventional, observational studies.
- To carry out the in-silico external validation aimed to evidence how the proposed AI tools can contribute to improve different aspects of the management of lung, colorectal, breast and prostate cancers. The clinical validators will be assisted by technical experts, in order to discriminate usability aspects of our AI tools (which refinement should be subject of future innovation projects and validated in prospective clinical trials) from performance issues which are this project´s core.
- To contribute to the technical, operational and ethical standards for AI-based solutions in healthcare.
WP10: CHAIMELEON Repository Sustainability after Project End
WP10 is devoted to the panning and activation of actions concerned with the sustainability of the repository after the project end. It is led by UNIPI. Its specific objectives are:
- To plan actions for ensuring CHAIMELEON repository’s sustainability in the long-term, with an ambitious vision, for a Repository that will incorporate new hospitals and types of cancer.
- To execute the planned actions, which will include actions for (i) promotion of the repository’s sustainability by design; (ii) by additional funding and (iii) by synergies with other existing biobanks/repositories (e.g. European Genome phenome archive, European Cancer Information System) and with on-going initiatives to create EU-wide repositories (e.g. BIGMEDILYTICS) and other projects funded by this call.
- To coordinate with the technical WP2 and WP3 for adopting protocols (e.g. pseudonymisation tools) to facilitate linkage with other biobanks and repositories (e.g. pathological, molecular, genetic data).
WP11: Communication, Dissemination & Exploitation Plan
WP11 aims to disseminate the CHAIMELEON results to the international scientific community, to publish findings in high-impact journals on AI for cancer managment, to raise awareness of the repository, and to identify and protect IPR generated in the project. The WP is led by EIBIR managing the dissemination and communication activities of the project in close collaboration with MAT who lead the project’s exploitation planning.
- To ensure efficient dissemination of CHAIMELEON activities, progress and achievements.
- To ensure optimal visibility and a wide outreach to all relevant stakeholders.
- To disseminate results to the scientific community.
- To identify and protect IPR and to support uptake and exploitation of project results.
WP12: Project Management
WP12 includes the coordination and administrative management of the project. It is concerend with the implementation and monitoring of the management procedures, oversees the project progress, and is responsible for the coordination of reporting procedures. This WP also includes data protection management carried out by UV. The WP is led by HULAFE. The spcific objectives are:
- To carry out the project management, including the ethical management, to manage and optimise the application of resources and to ensure that the contractual requirements are met.
- To ensure the project management procedures are implemented.
- To coordinate with and report to the European Commission (EC).