Copy
Accelerating the lab to market transition of AI tools for cancer management
Share Share
Tweet Tweet
Forward Forward

Welcome to the CHAIMELEON Newsletter


Table of contents

 

The CHAIMELEON Project

Creation of a pan-European repository of health imaging data for the development of AI-powered cancer management tools

To raise the quality of medical diagnostics and treatment, management of big data by artificial intelligence for medical imaging depends on a validated and safe repository of data. CHAIMELEON, a 4-year project funded under the Horizon Europe research and innovation programme, brings together 18 European partners, among them hospitals, universities, R&D centres and private research companies, to provide expert practice, design and experimentation. With access to an EU-wide dataset of oncology imaging, the partners hope to improve interpretation, extraction and exploitation of imaging processes related to the four most prevalent types of cancer worldwide: lung, breast, prostate and colorectal cancer.

The repository will be populated with a wealth of MR, CT and PET/CT imaging and clinical data (approx. 40,000 cases), representing several decades of patient oncological management, and artificial intelligence researchers will develop a multimodal analytical data engine for data manipulation and work toward quality control in terms of data de-identification, curation, annotation, integrity security and image harmonization. With compliance with EU ethical and safety regulations assured, the repository can then be validated for usability by artificial intelligence experts as well as effectiveness by clinical studies conducted by the partner hospitals.
CHAIMELEON, in addition to external independent partners, will then assess and validate its tools according to common clinical oncological processes in use to ensure successful outcomes in diagnosis and treatment, thus establishing a repository accessible across the EU, harmonised protocols and the infrastructure necessary to solidify the contribution of artificial intelligence in oncological care.

In this way, the project will provide a much-needed resource to accelerate the clinical translation of AI tools for cancer management and will open up a new horizon for AI-enabled, improved disease prediction, diagnosis, and follow-up. Beyond CHAIMELEON, the repository, its infrastructure and analysis tools will be adapted to expand to other types of cancer and potentially for other widespread disorders, from cardiovascular to neurological or psychiatric diseases.
Learn more

Interview with CHAIMELEON Coordinator Luis Martí-Bonmatí

How was the idea for CHAIMELEON developed? What immediate need does it address?
We knew about the reproducibility crisis from years as the main limitation for radiomics implementation in clinical practice. AI will probably harmonize real-world data and images allowing radiomics integration in the AI-developed prediction tools. CHAIMELEON was created with the aim to create an AI-powered multicenter secure repository to allow experimentation for the training and testing of cancer management AI tools. This large interoperable dataset meets the current need of unleashing open access to large, harmonized and quality-checked multicentric and multivendor sets of annotated imaging and related clinical data for the 4 most prevalent types of cancer worldwide: prostate, breast, lung, and colorectal cancer.
The CHAIMELEON concept was developed in the context of the fast-growing field of clinical AI tool development, which needs access to large and quality-controlled data sets for the training and testing of tools before they are clinically validated and released to the market.
 
What is the benefit of a European and interdisciplinary approach to tackle this need?
Data complexity and AI model developments need a large European consortium. Creating a large repository from different Electronic Health Records sources requires the coordination of clinical centers, IT experts, AI developers, data and radiomic scientists, as well as cybersecurity and legal specialists. The repository will be populated with cross-vendor/cross-institution image datasets. Since the reproducibility of Quantitative Imaging Biomarkers (QIBs) in radiomics is pivotal, CHAIMELEON will develop AI data and images harmonization approaches and their clinical validation as the main objectives.
 
What makes the CHAIMELEON repository unique? (In comparison to others)
The main differential aspect of CHAIMELEON is the refining and testing of the different image harmonization approaches needed to perform emulated clinical trials as observational studies on retrospective data in cancer. The cancer data and images repository is created as an AI-powered platform allowing AI experimentation on the cloud. An analytical data exploitation engine allows the extraction, interpretation, and exploitation of all the information stored at the repository. The decision support tool will help patents precisely evaluating their
situation and estimating tumor phenotype, treatment effect prediction, and time to event prognostication.

What are major challenges you expect to face and how will you address them?
We are facing difficulties related to ethical and legal restrictions for the secondary use of clinical data for research purposes, operational and technical complications related to data extraction from data provider sites and from undefined health data standardization processes. Another challenge is to find the best strategy to ensure the long-term sustainability and usability of the repository, while maintaining an open access resource to the research community.

How will the work in CHAIMELEON improve cancer management for clinicians and patients in Europe?
The creation of the CHAIMELEON platform will constitute the establishment of an open access data and tools resource, fostering AI experimentation in the development of cancer innovations. These tools are targeted to assisting radiologists and oncologist by predicting the precise tumor expression, allocate best treatment and predict time to relapse for lung, breast, colorectal and prostate cancers. We hope that the creation of this repository and platform will accelerate the lab-to-market transition of AI tools, facilitating their launch and deployment at hospital sites and hence enabling a faster, more accurate and more reliable disease management by clinicians. The in-silico evaluation of treatment effects will benefit patients. I hope that this will have a positive economic and social impact in current medical practice, untapping enormous potential towards new advancement in the field of cancer management.
 
On a personal note, what motivates you most in carrying out this project?
The creation of a fully multidisciplinary team, having the know-how on imaging, radiomics, data curation, repository hyper-architecture, observational studies and clinical trials has been an extremely positive experience. Being able to foster massive data extraction, understanding the common data models and interoperability standards, being able to construct observational studies and minimizing the impact of data instability are fundamental aspects of my motivation. Even more, paving the way towards a European Cancer Images initiative, allowing the deployment of AI solutions on large data and complex sets is the vision for the end of our project.
 

First Results

During the first year of CHAIMELEON, the partners laid the foundations for the design and set-up of the CHAIMELEON repository on a technical, clinical, and legal level.

The partners surveyed existing initiatives and established an Imaging Health Data Map that gives the project the opportunity to learn from best practices that will assist the consortium in designing and setting-up the CHAIMELEON repository. At the same time the legal grounds and all ethical and legal/regulatory principles were defined. Based on this, the interim design architecture to implement the CHAIMELEON repository was established: user roles, user stories, use cases, and requirements were collated, described, and analysed. This work will inform the further technical development of the repository but can also serve as general guidance for building a repository. As most of the components are released under open-source licenses, the guidance provided may assist researchers in employing the CHAIMELEON design for building repositories to address similar problems.

Moreover, the partners have prepared for data collection at the local level at each data provider site for their future upload onto the central CHAIMELEON repository defining the clinical objectives and related data to be collected, ensuring regulatory compliance, implementing tools for automatic extraction and curation of data from different sources, and enabling technical readiness for the deployment of these tools and connection to the hospital’s Picture Archiving and Communication System (PACS) and the different clinical data sources at the local level.

Read up on our first public results on our website:

Our Publications

 
Read up on our latest publications related to the CHAIMELEON project:
Read more
The CHAIMELEON consortium consists of 18 hospitals, universities, R&D centres, and private research companies from 10 European countries that constitute a pan-European ecosystem of knowledge, infrastructures, biobanks and technologies on oncology, AI/in-silico and cloud computing addressed to health.


CHAIMELEON will set up an EU-wide structured repository for health imaging data as an open source for AI experimentation in cancer management.
Visit the CHAIMELEON website
Twitter
LinkedIn
Website
Copyright © 2021 European Institute for Biomedical Imaging Research (EIBIR), All rights reserved.


Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list.

Email Marketing Powered by Mailchimp

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 952172.
www.chaimeleon.eu