FAQ - Cancer Image Europe

Which EU Member States and which clinical sites are currently involved in EUCAIM?

The EUCAIM website provides information on the clinical sites involved in EUCAIM project, with an overview per country, differentiated by provider type (research infrastructure, clinical centre, etc.). This information will be updated as new stakeholders are incorporated.

Will the federated infrastructure include histological images?

The infrastructure will address mainly medical images from devices such as for example MRI, CT, ultrasound, single photon emission computed tomography (SPECT), positron emission tomography (PET), X-ray, and mammography. In addition, the infrastructure is to be linked to other repositories, including biobanks for histological images, for example via the BBMRI-ERIC or the infrastructure created by the BIGPICTURE project.

How will the quality of data be controlled and ensured? We know that annotations vary from one institution to another. Will you have mechanisms to foster data quality?

It is an important challenge to curate the data in order to be able to develop robust AI models. EUCAIM has tools to monitor the image quality, in particular tools that can be used to annotate images through machine learning models, and tools to evaluate the quality of newly acquired images through a quality index score. Data quality criteria and data quality tools will be further elaborated by the project and data curation tools will be further improved. It is also expected that the infrastructure will help finding new self-learning methods to cluster the images and provide such services to data providers.

What is the role of proprietary data in the infrastructure created by EUCAIM project?

There are occasions when data is used that may be protected under a license, being proprietary to an organization or a project. All proprietary data will be identified in the data availability statement. The statement will further describe restrictions on reuse and federated use of this data. In the long term, EUCAIM will favour the storage of non-proprietary data within the Atlas of anonymised Cancer Images.

How will EUCAIM manage the legal requirements of the General Data Protection Regulation (GDPR) when collecting the data? In practical terms will data leave the hospital? How will you ensure data privacy?

This is a key standpoint for any European project based on unleashing access to health data. To ensure full compliance with data privacy principles, EUCAIM will be built under the data privacy by design and by default approach.

There are currently two scenarios for the implementation of the infrastructure: in the first one, data in the fragmented research repositories is already anonymised (e.g., data coming from fragmented repositories such as previously EU-funded projects and other research repositories). In this scenario, privacy is no longer an issue. Instead, the challenge will be how to make these repositories interoperable to be able to populate the Atlas of Cancer Images. Anonymized data from established research (such as data coming from clinical trials) will be easier to include, as this data is already pseudonymized and only a final de-identification step is needed (full anonymization).

In the second scenario, related to data at hospitals and real-world clinical practice, we are pursuing a federated learning approach, whereby the data remains distributed, stored at the local sites, with all the protective privacy measures in place. The data will not leave the hospital premises, safeguarding its privacy. As part of the project implementation, we will investigate how to do this securely and evaluate the associated risks. To allow this, the infrastructure software will get access to the data warehouse at the hospitals, for which the hospitals will need to have a structured data warehouse following the requirements that will be defined by the EUCAIM project. With data warehouses following these standards and the data being safe and quality-controlled, the process of extraction of the data from the Electronic Health Records (EHRs) to the data warehouse will be harmonised and normalised in an easier way. For this, a checklist will be provided by the project to the different hospitals, and if the hospital has enough computing resources, it may be possible to deploy AI solutions directly on the data in the hospital. The data won’t have to leave the hospital. The discussion on the best data warehouse architectures has already started. The project will publish these guidelines and checklist within the first year to provide technical guidance to individual hospitals.

Will clinical trial data from industry sponsored studies be included in the infrastructure? Are these data contributions from clinical trials already secured?

One of the objectives of the EUCAIM project is to support well-designed research on data, observational clinical studies and AI development trained and validated on a large number of images. Within the infrastructure, cases will be recruited from both real-world environment and completed clinical trials. Furthermore, the central hub will provide collaboration and support for the planning of observational clinical trials, targeting clinicians, researchers and innovators. EUCAIM work on technical AI developments on images and observational studies on data as emulated clinical trials will foster social innovation frameworks to produce practical insights in the real-world oncology practice. To do this, the infrastructure has incorporated and clustered not only technical and clinical partners, but also European Research Infrastructure Consortia (ERICs) and European coalitions dealing with patients, medical associations, academia, industry, enterprises, healthcare systems, regulatory bodies, and policy makers.

Once there is a solution, a model or a decision support system coming from observational study on retrospective data, this solution has to be validated in daily clinical practice. The best way to do that is through the initiation of clinical trials. It is within the scope of the project to define how these clinical trials have to be constructed. It is expected that clinical studies will then be initiated by pharmaceutical companies. EUCAIM stakeholders like the European infrastructure for translational medicine (EATRIS) will help on the definition of the clinical trials.

Countries from Central/Eastern Europe seem to be less represented in the EUCAIM consortium. Will there be opportunities for countries not represented yet to join the consortium? When?

The EUCAIM project will organise an Open Call to support the extension of the infrastructure, inviting new data providers and researchers to join. The Open Call for the federation of new data hubs will be launched, aimed among others at establishing longer-term collaboration with new cancer image data providers, including setting up the required technical infrastructure. The onboarding of new data providers is intended to increase the geographic dimensions, data modalities or cancer targets, and facilitate the uptake of new trustworthy AI algorithms trained on the data of the repository. This Open Call is expected to be published in 2024, with new data providers being onboarded in 2025. In addition, there will be an engagement strategy for new stakeholders to join the infrastructure from the beginning of the project.

Are there any clinical questions of particular importance that medical researchers would like to see addressed in EUCAIM?

Oncology is the main clinical aspect that will be addressed. There is no focus on specific pre-defined tumours or clinical questions. The project will deal with practical unmet clinical questions coming from researchers, practitioners and innovators, such as easy recognition of small lesions, phenotyping of tumours, dealing with prognosis and treatment allocation, and in the near future, evaluation of staging. The platform will be open to all types of tumours, including for example breast, lung, colorectal or skin cancer. Even though the platform will not initially focus on addressing specific clinical questions, it is planned that by project end a set of clinical use-cases will be chosen for platform testing and validation. This will allow fine-tuning of the different platform implementations in terms of requirements and capabilities.

Will there be any emphasis on childhood cancers and involving adolescents and young adults with cancer in research?

We aim for EUCAIM to become a cancer-neutral federated repository and to push the infrastructure to be able to solve clinical questions relating to all cancer types, including childhood cancers. The created infrastructure will be open to all cancer types and it will initially cover at least 5 cancer types: breast, lung, colorectal, prostate cancer, liver and paediatric tumours. The project builds among others on the results of the PRIMAGE project which was one of the biggest research projects on childhood cancer (neuroblastoma and Diffuse Intrinsic Pontine Glioma (DIPG)). PRIMAGE is one of the EUCAIM stakeholders and sharing cases and results is currently being discussed. The project will establish links to initiatives and projects working on paediatric cancers, such as the EU4Child project to explore synergies and links.

How will the Testing and Experimentation Facility for Health (TEF-Health) and EUCAIM be connected on the infrastructure level?

TEF-Health and EUCAIM represent two complementary pillars of the European Cancer Imaging Initiative. The design of EUCAIM will be technically compatible with the TEF-Health project (in terms of interoperability, standards, and data models). The EUCAIM consortium plans to closely collaborate with the TEF-Health project in order to ensure that TEF-Health is one of the processing back ends supported in EUCAIM. EUCAIM infrastructure follows a federated and distributed processing model, in which the TEF-Health could be considered as a supported backend for the processing of data. EUCAIM has defined a communication channel with TEF-Health to align the developments in both infrastructures.

How does the European Cancer Imaging Initiative ensure that there is (or will be) an active participation of the patients either as a stakeholder or a partner? How will patient representatives/organisations be involved in the EUCAIM project?

In addition to the different dissemination actions, the EUCAIM Engagement team will organise regular engagement meetings: including not only hospitals but also consultations with representatives from the main stakeholder groups: patients, data providers, service providers and data users in academia, industry and policy makers. The training activities will include activities targeting patients’ organizations to inform them on the secondary use of their health data for research and innovation (GDPR Compliance). Patients and patients’ organisations such as Patvocates and WECAN, the Workgroup of the 23 pan European cancer patient advocacy networks, will also be involved in the development of the Ethical, Legal and Social Implications (ELSI) data framework, for example they will be invited to participate in interviews and surveys. Moreover, the Stakeholder Advisory Board (SAB) which will be set up to gather and exchange additional feedback from stakeholders on the progress of the project. It will comprise representatives pertaining to the different partners and stakeholder groups in the remit of the project, including organisations representing patients such as the European Cancer Organization. Finally, the patients’ organisations will be closely involved in the work on data altruism.

How will you address the risk of participating institutions becoming more reluctant to share and annotate data as the time goes by/when the EU funding for the EUCAIM project is over? Do you have a policy in place to counter this risk?

At the outset, the EUCAIM project builds upon existing projects and research outcomes. It will link existing repositories and add new ones who are willing to share either metadata (catalogue) or data (images and related annotations). The risk of not having data is lower than in a standard research project, as EUCAIM builds on the AI4HI network of research projects. Since the beginning, EUCAIM will also work with the hospitals, to check on warehouses quality and support the hospitals in structuring their work to allow them to do observational clinical studies. The project will provide a seamless onboarding process and technical and legal support, including annotation tools to data providers. We expect that new hospitals that are not part of the consortium will see value in participating to maintain and ensure their own competitiveness on the national and European level as regards the participation in research and innovation projects and the construction of evidence-based tools ready to be brought to the market.

Will you engage with standardisation organisations to improve and adapt existing standards with project knowledge?

Standardisation and normalisation will be essential to the project. EUCAIM builds on the creation of similar projects which have already put an elaborated list of standards in place, particularly pertaining to interoperability standards for the infrastructure and data harmonization and standardization amongst others. The consortium is formed by experts in this field, and the project has a dedicated work package ensuring these standards are put in place and monitored throughout the project lifespan. Information technology standards on interoperability (such as Open ID, RESTful services, secure transport protocols), de-facto standards on dataset cataloguing and data representation, and standard file formats (such as DICOM) will be considered in the implementation of the infrastructure. The Spanish Association for Normalization (UNE) is also a stakeholder of EUCAIM and will provide consultancy and support on all the standards and norms that apply to the project.

For the sustainability and impact of the new services and infrastructure, it will be of interest to develop training solutions tailored to the new services (targeting health care professionals as end users). How do you envision the engagement of academia (universities, polytechnical schools) in this step?

As part of the project implementation, EUCAIM will create an Engagement team responsible for making connections with all types of relevant stakeholders, including not only start-ups, but also technical universities and hospitals. They will not only be actively contacted by the team but also regularly informed about the project progress, so that opportunities for involvement and collaboration can be created as the project evolves.

Is the European Cancer Imaging Initiative limited to AI or do you accept also other imaging technologies as hardware, software and firmware for time-of-flight PET for early cancer detection and low dose usage of FDG?

EUCAIM is a data infrastructure enabling a wide span of technologies, therefore is not limited to AI and is fully open to any technological improvement, including image acquisition. The scope of the Atlas of Cancer Images will be enriched by adding new data from new devices and protocols, as EUCAIM will stand as the platform for research and innovation on medical imaging by using any device and methodology.

Is there a plan to develop a standard content of the data to be made available to researchers? In particular, is data on radiation exposure to be included e.g., to allow developments around meeting EU legal requirements (Euratom Directive) for radiation protection of patients?

Standards will be used all over the platform. Partners related to standardization are within the EUCAIM Infrastructure. Standardized and structured DICOM data on radiation exposure will be incorporated and available within the platform.

Will Cardio Oncology Imaging of the heart be included in the Cancer Imaging and AI initiative? Cardiotoxicity and Heart Failure is one of the potential side effects of chemotherapy treatment of cancers.

EUCAIM is a pan-cancer Infrastructure, not limited to any specific cancer types. As such, images from all patients having all types of treatments will be incorporated. It is expected that the Atlas of Medical Images will also allow for the grading and estimation of toxicity and late side events, including cardiotoxicity.

How can a start-up with a proven technology to early detect cancer be part of this Initiative?

The best approach would be to first contact the to identify synergies and potential areas of collaboration. EUCAIM aims to use knowledge already generated and integrate already viable solutions, so interested parties with a solution in place may contribute by sharing their previous experience and expertise. EUCAIM is favouring start-up engagement through both and participating in our Stakeholder Board. The datasets available in EUCAIM will be organized considering regulatory practices, to facilitate the certification of new medical devices, by allowing the start-ups to develop clinical studies generating results ready for CE mark submission under the new Medical Device Regulation (MDR).

Questions from the Technical Webinar (March 14, 2024)

Regarding the (future) cancer imaging database: how will it differ - or be similar to - the cancer imaging archive established in the US?

We are focusing on supporting different scenarios for data access, including distributed processing and federated learning, avoiding data from being moved from the data holders to the central storages, and in-situ processing of data deposited in the reference sites, so data is not downloaded, and traceability is enhanced.

Not being able to download the data may be an obstacle for some companies

We are working with the data holders to define data licences that will permit downloading. We aim to end up with these three access models (federated, in-situ, downloadable).

Can you describe the data governance?

Data access will depend on the licence, and we foresee different access models, depending on the conditions. Data holders will participate in the data access decision.

Who is in charge of quality control of the images?

We will make quality analysis tools available to the data holders so they can assess the quality of the data. An access committee in EUCAIM will check the features of the data, and finally, we plan to add a community evaluation through ranking.

What different imaging formats will be supported? Will you use open formats for machine learning such as onnx.ai?

We are currently supporting DICOM and planning to support other formats in the future.

How easily did/do all patients provide their consent to use their images for AI?

The consent is given throughout the institutions. The project will also prototype a Data Altruism activity in which citizens will provide their consent for the secondary use of their data.

Where is the central cloud is located?

We have two central (reference) nodes. One in Spain and one in the Netherlands.

What are the benefits for the data holder?

Data holders will benefit of participating in a research network. Direct benefit to researchers within the data holders will be the academic recognition through the citation of their datasets (it is expected that new frameworks such as COARA – https://coara.eu/ will foster the recognition of open science activities. Centres can benefit by being members of further observational clinical trials, which will have revenues for patient recruitment.

Is there any financial support for data assembly, annotation and provision?

Yes, we have an open call to be published soon. As a first step to expand our data we will host an open call for data providers (and AI tool providers). Funding is available for successful applicants. This can be up to 214,000 Euro, however with a 50% funding rate – meaning that the European Commission/the project will contribute half of this total sum, and the successful applicant will have to cover the remaining sum from other sources.

What exact technical solution is used for federated node?

We are developing a solution based on Flower. This will be available in June. You can find more technical information in the architecture document.

DICOM is quite “diverse”. Are you building your own image readers?

We have integrated DICOM viewers with reference storage sites. We’ll show during the demonstration.

How many and what kind of imaging data are available currently?

We are focused on MRI, CR, mammography, PET and other data.

What kind of tools are made available?

In the first phase, mostly data preparation, quality assessment and annotation tools. The platform will be open to any relevant tools that Software provider may offer.

What is the usual time between the data access request and the (potential) positive answer?

We are building this currently. Technically this is solved but we are now in the process of signature of agreements. I cannot answer by now.

What are the benefits for my employer? I will have to justify any involvement in the project as being beneficial to the hospital.

Similar to the previous questions on Data Holder. Recognition throughout the consideration of data as first-class research objects, networking within further clinical trials, adherence to Open Science principles, etc.

Many hospitals and research centers are usually reluctant to share their data. Do you think this is a challenge? If so, how do you foresee to overcome this challenge?

We are intensively working on that. This is the reason of having different access modes (including federated). We are working on data licences that could reduce the reluctance

Will there be cloud-based compute resources available - or how shall federated learning be harmonized across centers?

Yes there will be. Also a federated node will need either local or cloud-based compute.

How are communities registered?

Our recommendation is to register through projects or umbrella organisations as stakeholders. We will be supporting working spaces for communities through agreements.

Is it possible for industry partners to get access?

Yes this is possible. For data access it depends on whether the datasets allows it.

Which ontologies do you use for annotations?

We are using DICOM Seg for the segmentation. The full ontology is defined in https://zenodo.org/records/10817687

Do you have the average age in addition to the low/high age?

It can be calculated by the platform and it can be considered as a recommendable field. This will be proposed to the Data ontology activity leaders.

All datasets in public catalogue require later registering?

No, it also allows for public datasets that are available to all users.

Are access requests already available? Or when would it be possible?

Access request is pending for the finalisation of the agreements and will come soon. The technical procedures are completed.

Alongside the imaging data it is often useful to have the associated pathology data for final diagnosis and understanding lesion measurements at the pathology level. Is this data also available to complement the imaging data?

The datasets have not only images but also the associated clinical data. The full ontology is defined in https://zenodo.org/records/10817687

Additionally for CT scans would this include both native and contrast enhanced images for the same patient so they can be analysed in parallel?

We are using RWD (Real World Data) so it depends if the data holder has provided it.

Is there a fee for users to EUCAIM?

We are defining different usage models for the sustainability of the platform. It will depend on the resources requested but there will be services provided at zero cost.

I do not understand why you have to evaluate the scientific value of a proposed project from somebody that want to use the dataset?

The access has to be bound to a purpose that fits with the Data Sharing Agreements signed with the Data Holder. We have to ensure that the use of the data is scientifically and ethically sound. Moreover, we are providing computing and storage resources through the platform and we have to ensure a good usage of them.

In the demo, Ignacio had to sign in in his local university portal. If our institution is not added to your platform, how is the process? I guess if we would like to be added we should apply to the open call in April right?

Other providers such as Google accounts will be permitted

Besides data, you mention that you want to add "AI tools", what do you mean by that? Trained models, code? What would be the advantage of using your platform instead of already free and open platforms as GitHub, Huggingface, etc.?

We want to address both AI mode training and AI model usage. The advantage with respect to platforms as HuggingFace is that we keep the data within our secure platform.

The requested information to access data is more oriented to a research study. Is there another scenario for developing AI models for commercial use by SMEs or large companies?

Commercial use will be supported. We are working on the legal framework to support it.

This is a great tool! Do you think it could be used as a basis for the UNCAN.eu platform?

We are in touch with UNCAN.eu to see how we can best link and integrate our developments.

Do you offer storage for scientists doing the analysis during the duration of the investigation? If so, how much storage? How many concurrent users does the platform support (when doing analysis)

Yes, it will depend on the objective of the study (it can be required at the proposal application). We have tested the platform services with 30 concurrent users using a small fraction of the resources (less than a quarter) and worked fine. Expecting to support on the order of 100 concurrent users.

Can user create their own virtual environment e.g. conda env to use own library?

A user can create custom containers and run them using a user-level container system named udocker (https://indigo-dc.github.io/udocker/). Instructions will be available.

Can individual researchers, small start-ups or NGOs request access? Or must users belong to a more established institution (university, research center, companies..)

Yes they can. However, an institution account helps for authorisation, but is not required.

Which compatibility level are we working on right now? I mean in these examples

The examples shown are in tier 2 (limited). Tier 3 will arrive by July 24.

Can the user view the image before requesting access?

No this would not be possible. Collection characteristics in the public catalogue and federated search is available for requesting access.

Can a data holder limit data access? be in control of how and who accesses data?

Data holders can express the access conditions in their Data Sharing / Transfer Agreements, which must be aligned with the EHDS regulations.

How does the publication process work? The data access regulation is very clear, but does the source of the data need to merely acknowledged or are individuals included in the author list? Can the data provider perhaps decide this?

This will be agreed between the data provider and the EUCAIM platform in the collaboration agreement. Different options are possible.

Can this software use globally for tele-radiology service also?

The software is intended for secondary use of data (i.e. research, quality evaluation etc, software development) and not for use in clinical practice (primary use)

Are you possibly envisioning linking clinical data to the images? How are you addressing this, if so?

The datasets have not only images but also the associated clinical data. The full ontology is defined in https://zenodo.org/records/10817687

If data are downloaded for local AI model training — what are the “rules” for storing EUCAIM data locally? I see the data are subject to GDPR

At this moment we do not allow data downloading. We are working on Data Transfer Agreements to support it.

Who will hold intellectual property on the the tools developed within the platform?

IPRs are always property of the developer. EUCAIM expects to negotiate advantageous usage conditions of the foreground developed within the platform.

I've recently applied, but my application was refused due to the platform being still experimental. Has that changed? Should I re-apply?

Please wait until we set up the new servers. It will be available in the short time. Thanks for your interest.

Are you using Secure Multi Party Computation (SMPC)?

The federated framework will be based on Flower.

What is the viewer you have shown with CHAIMELEON data? it is free?

Not at this moment, but we are working on exposing it to external users. Hopefully this will come in weeks.

Regarding federated concept, is there any support (either technical or financial) to locally collect/anonymise/curate/collate datasets that can then be shared as a federated provider within EUCAIM?

We have the Open Call and we also have Technical and Support committees defined to support the federation.

Does EUCAIM support documents preparation for ethical approval?

Not at this moment. We can work on that if necessary.

How the National and Regional Level Healthcare Authorities will participate?

They can participate through the hospitals involved or through their research networks. We can provide examples on how our Data holders contribute.

Do you also need histopathological findings or just purely pictorial information?

The platform is open to this, but not required.

Do you need manual segmentation of tumors and other tissues?

Yes this highly prefered.

We are currently planning to make an MRI dataset available in the near future. On the page of ‘Become a Data Holder of EUCAIM’ I don’t see concrete first steps/contact points. We are currently preparing the documents for the ethical committees, but the platform to share the data should be mentioned in there. Is EUCAIM already going to be available in a short time span?

Yes, June 24 will be ready for admitting projects to upload data. We will provide support such as dedicated research community space and the ability to work on the datasets prior to share them on the platform.

Feed should be linked to sustainability and perhaps whether user is commercial or public?

We want to address both public and commercial usage, but ensuring that the restrictions of the data holders are clearly met.

In the case of MRI, is it preferable to have research sequences (diffusion spectrum imaging etc.) or is it better to have standard sequences?

It will depend on the objective of the study. We are supporting different types of data protocols.

How do you make sure that the data provided is adequate quality?

The evaluation of the quality is a responsibility of the data holder but it will be evaluated in terms of coherence and adherence to the EUCAIM model by the project during the quarantine period, using a set of tools. Evaluation of users is also considered.

In case of CT it is advisable to have standard data or scientific data, e.g. dual energy CT (calcium suppression, Z effective...)

Please check the full ontology of EUCAIM in https://zenodo.org/records/10817687

Is there a minimum of data or cases that need to be uploaded to be a provider? Do the cases need to be uploaded one-by-one or as a batch? Which are the format specifications for data (imaging and clinical)? Can longitudinal data be uploaded?

We do not have a minimum threshold as it may depend on the type of data and its quality. This will be evaluated by the scientific committee. We will provide batch uploading. Longitudinal data is under consideration but we do not have a final decision taken.

What are the technical requirements to become a node of the federated network?

Details are published in the D4.3 Rules for participation deliverable. It will be available in our website shortly.

Can the datasets in EUCAIM be used for challenges?

Yes! this was one of the cases we experimented in one of the collaboration project. However, this has to be well-planned, as Challenges consume a large set of resources.

Can we use data we have already published in scientific papers? Can there be a requirement to cite these publications to provide data?

Yes and we want to promote datasets as first-class research objects, providing a doi for each one.

Questions From The Technical Webinar (April 26, 2024)

Please note that some questions were answered live in the webinar, therefore we encourage you towatch the recording of the webinar, if you have not attended.

Does the Cancer Imaging Initiative also include Microsopic imaging (eg. clinical pathology)? This would be very interesting for us as Innovator?

At this moment we are focusing on Medical Imaging modalieties such as CT, MRI, PET, etc. However, some of our partners and stakeholders address these type of image data, so I would suggest contacting them.

t would be usefull to know what exactly are the imaging modalities already available in this data repository , e.g. MRI, CT, US…. Also, how many datasets are available per modality, per type of cancer, organ etc…

You can browse our public catalogue for an overview of available data: https://catalogue.eucaim.cancerimage.eu/#/

This is kept up-to-date, so as soon as more data is available, this will automatically be visible in the catalogue.

Does the Cancer Imaging Initiative include MR Spectroscopy ?

We are open to receive this nodality, although we do not have cases yet.

Has the EUCAM Initiative an interest in involving European research teams who design free and open-source software dedicated to cancer imaging?

Absolutely! Very happy to get those teams involved.

Is there a minimum number of subjects required for the new data providers applying on june 10th call ?

There is no threshold, as some cancer types will be more challenging. A small but representative number in a rare disease could be more valuable than a large number on a common one

All Cancer Types are OK for the call ?

Yes, we accept all cancer types.

Do patients need to sign informed consent specifcically for this platform?

This most likely depends on the legal and ethical legislation in your country. For some countries we are aware that such secondary use within the EUCAIM platform would require informed consent specifically for this platform, whereas in others it is not explicitly required.

What requirements must datasets meet? Should they be labeled?

Mainly we expect DICOM data with a minimum annotation and if possible the segmentation of the regions of interest.

Who is the owner of IP for the AP algorithlms developped by federated learning on data from this repository?

The IPRs of the tools developed within the platform are owned by the developer, but EUCAIM expects to have the right to provide them to other users of the platform.

In the call is described that "Applicants must commit to at least 50% co-funding of their maximum total budget". Do we have to put that 50% of the budget to participate in the project?

Yes, the 50% funding rate applies for this call. We will explain this a little later in the presentation.

We are a SME focused in AI. Could we participate in the Open Call tohgether with a clinical institution and incorporate their own data to the platform to use in our research?

Yes, that would be possible

How does EUCAIM relate to the US-based TCIA (The Cancer Imaging Archive - https://www.cancerimagingarchive.net/)?

We are providing a secure processing environment, with distributed processing capabilities, along with the data.

Is the planned output of the EUCAIM project also the creation of a set of metadata (defined using interoperable terminologies) for the description of whole slide images?

Yes, indeed. We are working on a hyperontology. The Beta version is in https://zenodo.org/records/10817687

Are PET/CT imaging studies of interest?

Yes, indeed.

How are the data annotated/labelled? demographic and clinical data only or tumor segmentation as well? Who does this annotation and how is reproducibility assessed?

We have some recommendations, and a hyperontology defined. We will provide some guidelines and tools for assessing the quality.

are there specific cancer types that the consortium is in greater need than other?

You can have a look at the types of cancer that we already have in the platform here: https://catalogue.eucaim.cancerimage.eu/#/ Of course we welcome cancer types not yet included, but this is not at all an requirement.

We are already involved in ChAImeleon project so already data provider for EUCAIM?

CHAIMELEON will be 100% compatible with EUCAIM so transition will be seamlessly implemented. It will be a question of singing the proper Data Transfer/sharing agreements.

As a data holder, is it possible to apply with data for multiple use cases (e.g. breast cancer dataset, gastric cancer dataset etc.)? And if yes, is there need to fill in the application form for each use case or can we bundle multiple use cases in one form?

Understanding the splitting among cancer types will be valuable for our evaluation.

As a data holder, is it possible to apply with data for multiple use cases (e.g. breast cancer dataset, gastric cancer dataset etc.)? And if yes, is there need to fill in the application form for each use case or can we bundle multiple use cases in one form?

As Ignacio said, the split would be interesting – but we’re of course interested in applications that provide data on multiple cancers.

Is it possible to amend / negotiate the data transfer agreement as applicable, e.g. in case there are specific requirements by a conbtributiong institution?

We expect that the DSA/DTA will need to be customized to the specificities of each centre. However, the spirit of the agreement and the compliance with the EHDS regulations will be mandatory.

You are looking for data and AI algorithms, but it looks you already decided to use FedBioMed and flower as FL infrastructures. The call can cover also proposal of other FL infrastructures or a bechmarking of them?

The frameworks have been defined by the group in charge of the distributed processing. Unless there is a strong reason, I do not think that other frameworks, unless they can fit in our distributed processing environment.

Re: “Understanding the splitting among cancer types will be valuable for our evaluation.” —> Can this be done in ONE single application form? Or do you recommend multiple forms, one for each cancer type?

You can submit this as one application, or as multiple – however you think it best represents the work you propose.

What is the current proportion of cancer types and kind of imagery already available?

Please have a look at our public catalogue: https://catalogue.eucaim.cancerimage.eu/#/

If we use this registry for developping, testing and/or validationg an AI algorithm that would further be cleared as a product for clinical use (EU-MDR and/or FDA), would the data be availble for regulatory bodies, in case of audit? How long is the archiving guaranteed?

This is our intention. We are working on defining the way to support industry on the certification of their products. Data will be FAIR (Findable, Accessible, Interoperable and Reusable) so references will persist.

Is there a minimum number of subject images that is required in order to participate in the call?

There is no threshold AFAIK, as some cancer types will be more challenging. A small but representative number in a rare disease could be more valuable than a large number on a common one.

are you only allowed to use for example prostate cancer MRI cases if you applied for this specific use case? for example if you are also interested in liver mri along the way are you also allowed to use these cases?

You could request access to multiple datasets that comprise different image modalities (different subjets too). If the dataset you request has different types of data, you will be able to access all of them.

We are a SME, which is part of a larger international group. Can we still apply to EUCAIM?

Yes. However if the international group is based outside EU you should check the elegibility conditions of Digital Europe.

Are compagies that are part of a big group elligible?

In principle, if they comply with all eligiblity criteria, yes. Feel free to reach out with more details to [email protected] to discuss this.

what is a PIC number?

It’s the number that your institution receives from the European Commission when registering in the funding and tenders portal to apply for funding or to manage projects that are funded through the EC: https://ec.europa.eu/info/funding-tenders/opportunities/portal/screen/how-to-participate/participant-register

Regarding funding, is it possible to provide it as personnel commitment?

You can provide any type of co-funding that your institution would allow in this case of projects. In-kind contribution through labour of your staff is typically eligible.

what is a PIC number?

Regarding funding, is it possible to provide it as personnel commitment?

You can provide any type of co-funding that your institution would allow in this case of projects. In-kind contribution through labour of your staff is typically eligible.

Are there any possibilities at all for UK organisations to collaborate, even if they do not receive any funding under Digital Europe? I believe that the UK is considered a non-associated third country.

Yes, I believe you’re right. See here: https://www.ukro.ac.uk/resources/eu-uk-relationship/

Are there any possibilities at all for UK organisations to collaborate, even if they do not receive any funding under Digital Europe? I believe that the UK is considered a non-associated third country.

I suggest that you join as a stakeholder. We will open collaboration for external partners, and being a stakeholder you will receive the information timely.

Can a research institute working on AI apply together with a data provider?

This is possible to an extent. Feel free to reach out by email at [email protected]. Please note that all applicants of course need to be eligible in this case. The overall funding would also remain the same – it does not increase if multiple institutions apply.

For ethical approval the work program description will need to be submitted in our Country . Can this be provided to us?

If you are referring to the Work Programme description in the Digital Europe Programme, You can find this here: https://digital-strategy.ec.europa.eu/en/activities/work-programmes-digital

As a data user, what happens with our algorithm? Are any rights surrendered?

I think you refer to the IPRs of an algorithm developed by you. If this is the case, the IPR remains on your side, although EUCAIM expects to have the right to exploit the tool by exposing it in its catalogue for the users of EUCAIM’s platform.

Is there any document or description on how you will monitor how the grant money will be used and what documents to prove it you will require?

Regular financial reporting towards the European Commission is required. This follows the usual procedures as described in the Grant Agreement with the European Commission. You can have a look at the model grant agreement that is referenced in the open call text.

How about data provisions that are exempt from filing for ethics approval, e.g. retrospective anonymized data?

We’re happy to receive such data as well – fully anonymized data of course requires less or no ethics approvals.

We are a micro company with focus mainly on developing privacy by design and default solution, and with passion for developing educational and training courses for privacy awareness. Is in the scope of this open call developing requirements for education of a Researchers, clinical scientific, and other stakeholders regarding risks noted in data protection impact assessment realted to developing, testinng and using AI algoritams?

This is very interesting, however the open call is limited to use cases that will allow us to test and validate the platform. Nevertheless, we would encourage you to join our stakeholder forum to engage and interact with the consortium and other stakeholders.

Good morning. Regarding the compliance with the DFF, is it mandatory to fulfill any of the three Tiers to apply as a federated node? Or the Tiers are the aim to achieve with the funds?

Federated nodes must reach the Tier 2 (all data is searchable) level and provide ways to access the data. Tier 3 means that you provide computing resources to run algorithms within your premises.

Can datasets also include non-EU data if provided through EU institution?

Most likely this is possible, provided there is ethics approval for the provision/use of this data from an EU-based ethics committee.

We would remove private data elements in the images (because these could contain sensitive data) even if they contain raw data that theoretically may be useful for processing?

Yes, we are facing similar problems too. Please check that you include the mandatory fields.

do you have any TRL preference when evaluating the proposal?

No threshold is defined. It will depend on the model target. In the case of innovative tools we will expect lower TRLs than tools for known problems.

Frequently Asked Questions

Which EU Member States and which clinical sites are currently involved in EUCAIM?

Will the federated infrastructure include histological images?

How will the quality of data be controlled and ensured? We know that annotations vary from one institution to another. Will you have mechanisms to foster data quality?

What is the role of proprietary data in the infrastructure created by EUCAIM project?

How will EUCAIM manage the legal requirements of the General Data Protection Regulation (GDPR) when collecting the data? In practical terms will data leave the hospital? How will you ensure data privacy?

Will clinical trial data from industry sponsored studies be included in the infrastructure? Are these data contributions from clinical trials already secured?

Countries from Central/Eastern Europe seem to be less represented in the EUCAIM consortium. Will there be opportunities for countries not represented yet to join the consortium? When?

Are there any clinical questions of particular importance that medical researchers would like to see addressed in EUCAIM?

Will there be any emphasis on childhood cancers and involving adolescents and young adults with cancer in research?

How will the Testing and Experimentation Facility for Health (TEF-Health) and EUCAIM be connected on the infrastructure level?

How does the European Cancer Imaging Initiative ensure that there is (or will be) an active participation of the patients either as a stakeholder or a partner? How will patient representatives/organisations be involved in the EUCAIM project?

How will you address the risk of participating institutions becoming more reluctant to share and annotate data as the time goes by/when the EU funding for the EUCAIM project is over? Do you have a policy in place to counter this risk?

Will you engage with standardisation organisations to improve and adapt existing standards with project knowledge?

Is the European Cancer Imaging Initiative limited to AI or do you accept also other imaging technologies as hardware, software and firmware for time-of-flight PET for early cancer detection and low dose usage of FDG?

Is there a plan to develop a standard content of the data to be made available to researchers? In particular, is data on radiation exposure to be included e.g., to allow developments around meeting EU legal requirements (Euratom Directive) for radiation protection of patients?

Will Cardio Oncology Imaging of the heart be included in the Cancer Imaging and AI initiative? Cardiotoxicity and Heart Failure is one of the potential side effects of chemotherapy treatment of cancers.

How can a start-up with a proven technology to early detect cancer be part of this Initiative?

Questions from the Technical Webinar (March 14, 2024)

Regarding the (future) cancer imaging database: how will it differ - or be similar to - the cancer imaging archive established in the US?

Not being able to download the data may be an obstacle for some companies

Can you describe the data governance?

Who is in charge of quality control of the images?

What different imaging formats will be supported? Will you use open formats for machine learning such as onnx.ai?

How easily did/do all patients provide their consent to use their images for AI?

Where is the central cloud is located?

What are the benefits for the data holder?

Is there any financial support for data assembly, annotation and provision?

What exact technical solution is used for federated node?

DICOM is quite “diverse”. Are you building your own image readers?

How many and what kind of imaging data are available currently?

What kind of tools are made available?

What is the usual time between the data access request and the (potential) positive answer?

What are the benefits for my employer? I will have to justify any involvement in the project as being beneficial to the hospital.

Many hospitals and research centers are usually reluctant to share their data. Do you think this is a challenge? If so, how do you foresee to overcome this challenge?

Will there be cloud-based compute resources available - or how shall federated learning be harmonized across centers?

How are communities registered?

Is it possible for industry partners to get access?

Which ontologies do you use for annotations?

Do you have the average age in addition to the low/high age?

All datasets in public catalogue require later registering?

Are access requests already available? Or when would it be possible?

Alongside the imaging data it is often useful to have the associated pathology data for final diagnosis and understanding lesion measurements at the pathology level. Is this data also available to complement the imaging data?

Additionally for CT scans would this include both native and contrast enhanced images for the same patient so they can be analysed in parallel?

Is there a fee for users to EUCAIM?

I do not understand why you have to evaluate the scientific value of a proposed project from somebody that want to use the dataset?

In the demo, Ignacio had to sign in in his local university portal. If our institution is not added to your platform, how is the process? I guess if we would like to be added we should apply to the open call in April right?

Besides data, you mention that you want to add "AI tools", what do you mean by that? Trained models, code? What would be the advantage of using your platform instead of already free and open platforms as GitHub, Huggingface, etc.?

The requested information to access data is more oriented to a research study. Is there another scenario for developing AI models for commercial use by SMEs or large companies?

This is a great tool! Do you think it could be used as a basis for the UNCAN.eu platform?

Do you offer storage for scientists doing the analysis during the duration of the investigation? If so, how much storage? How many concurrent users does the platform support (when doing analysis)

Can user create their own virtual environment e.g. conda env to use own library?

Can individual researchers, small start-ups or NGOs request access? Or must users belong to a more established institution (university, research center, companies..)

Which compatibility level are we working on right now? I mean in these examples

Can the user view the image before requesting access?

Can a data holder limit data access? be in control of how and who accesses data?

How does the publication process work? The data access regulation is very clear, but does the source of the data need to merely acknowledged or are individuals included in the author list? Can the data provider perhaps decide this?

Can this software use globally for tele-radiology service also?

Are you possibly envisioning linking clinical data to the images? How are you addressing this, if so?

If data are downloaded for local AI model training — what are the “rules” for storing EUCAIM data locally? I see the data are subject to GDPR

Who will hold intellectual property on the the tools developed within the platform?

I've recently applied, but my application was refused due to the platform being still experimental. Has that changed? Should I re-apply?

Are you using Secure Multi Party Computation (SMPC)?

What is the viewer you have shown with CHAIMELEON data? it is free?

Regarding federated concept, is there any support (either technical or financial) to locally collect/anonymise/curate/collate datasets that can then be shared as a federated provider within EUCAIM?

Does EUCAIM support documents preparation for ethical approval?

How the National and Regional Level Healthcare Authorities will participate?

Do you also need histopathological findings or just purely pictorial information?

Do you need manual segmentation of tumors and other tissues?

Feed should be linked to sustainability and perhaps whether user is commercial or public?

In the case of MRI, is it preferable to have research sequences (diffusion spectrum imaging etc.) or is it better to have standard sequences?

How do you make sure that the data provided is adequate quality?

In case of CT it is advisable to have standard data or scientific data, e.g. dual energy CT (calcium suppression, Z effective...)

Is there a minimum of data or cases that need to be uploaded to be a provider? Do the cases need to be uploaded one-by-one or as a batch? Which are the format specifications for data (imaging and clinical)? Can longitudinal data be uploaded?

What are the technical requirements to become a node of the federated network?

Can the datasets in EUCAIM be used for challenges?

Can we use data we have already published in scientific papers? Can there be a requirement to cite these publications to provide data?

Questions From The Technical Webinar (April 26, 2024)

Does the Cancer Imaging Initiative also include Microsopic imaging (eg. clinical pathology)? This would be very interesting for us as Innovator?

t would be usefull to know what exactly are the imaging modalities already available in this data repository , e.g. MRI, CT, US…. Also, how many datasets are available per modality, per type of cancer, organ etc…

Our open Call for new collaborators
launches in April 2024

SAVE THE DATE
March 14, 10:00-11:30 aM CET