The INCISIVE project launches the final prototype of the AI decision support toolbox and federated data repository to improve cancer diagnosis and care

Published on 06/08/2024

 

The INCISIVE project was officially concluded, receiving very positive feedback for its results from the European Commission project officer and reviewers. The project delivered the final prototype of an integrated platform with all planned functionalities, making it available to data providers, data users and healthcare professionals, following appropriate authorisation.

“The joint efforts and excellence of the multidisciplinary team of experts involved in this project have made it possible to successfully achieve our two main goals: an AI toolbox comprising AI services that support more effective decision-making for healthcare professionals and an interoperable federated data repository of FAIR anonymised cancer images and accompanying clinical data, in compliance with the GDPR” said the project coordinator, Gianna Tsakou of Maggioli SpA.

The project officer and reviewers concluded in their report that “many strong and pertinent exploitable results have been created (…). Key assets such as the AI toolbox platform and the federated imaging repository are highlighted for their exploitation potential”. The reviewers also acknowledged that, expectedly, certain challenges still exist in relation to the AI-based tools and engaging new data providers and stressed that “the project’s ambitious nature and numerous met KPIs demonstrate its potential for substantial scientific, technical, commercial, and societal impacts”.

The next big step is supporting interoperability, at all levels, among INCISIVE and other existing health data repositories and research infrastructures. INCISIVE partners are already working towards this goal under the Cancer Image Europe initiative. The active involvement of key INCISIVE partners in this initiative also provides a path for the sustainability and deployment of all major INCISIVE results, as well as an opportunity for the further training and validation of INCISIVE’s AI services as a step towards bringing them closer to clinical practice, as the reviewers suggested.

 

AI toolbox

The final prototype of the AI toolbox provides a set of explainable AI services and pipelines that healthcare professionals can use as decision-support tools with positive impact on the clinical workflow. It comprises a total of 28 AI models for lung, colorectal, breast and prostate cancer. The main decision-support services relate to the classification of abnormalities, patient prioritization, lesion segmentation and localization assistance, cancer diagnosis and staging and risk for metastasis prediction.

The input data to the INCISIVE AI services include imaging examinations (MMG, MRI, US, CT, PET-CT and histopathological images), and clinical metadata (patient demographics, biopsy results, laboratory examination results, tumor and treatment details). The INCISIVE AI toolbox integrates methods and components for explainability of the AI services so that healthcare professionals understand the reasons for the recommendations provided by the AI tools. Moreover, the toolbox integrates components supporting the usability of the AI services, such as image-to-report transformation and Augmented Reality visualization functionalities, thus facilitating the delivery of the AI services’ outcomes to healthcare professionals.

The usability and effectiveness of these services, and the viability and practicality of incorporating them into clinical settings were positively assessed by healthcare professionals in a pilot study and a feasibility study conducted during the project. These studies also provided valuable feedback towards the further improvement of the services.

 

Federated data repository

The INCISIVE federated data repository supports the data sharing of more than 5.5 million anonymized interoperable cancer images and accompanying clinical data from more than 11.000 individuals. All this data is coming from 9 distributed data providers that, in turn, have collected data from 14 clinical centres. The repository allows health data use among registered stakeholders in compliance with legal, ethical, privacy, and security requirements for AI-related training and research experimentation.

The repository relies on federated data storage and management, abiding to the highest data privacy and security standards: data holders are free to choose among setting up their own local data node that will interoperate with the platform or using INCISIVE’s central data node for secure data storage and management. The repository is also supported by data sharing mechanisms, covering both technical and operational aspects of data sharing, and complying with legal and ethical norms before data is shared.

 

Additional project outcomes

Through the development of the AI toolbox and the repository, INCISIVE partners also developed a federated learning mechanism that allows the training of AI models by leveraging distributed data stored in different locations, and a data interoperability framework. This framework includes a methodology for data integration, as well as a documented DICOM- and FHIR-based Common Data Model and tools supporting the integration of heterogeneous cancer image and clinical data from multiple data providers.

The project also developed a GDPR-compliant data sharing mechanism, covering both technical and operational aspects of data sharing. This mechanism includes data preparation tools and mechanisms that support data providers in sharing health data in a GDPR-compliant way. For example, it offers tools for data anonymisation, data collection, curation, annotation, quality checking, and more.

The INCISIVE consortium was coordinated by Maggioli SpA and comprised 26 partners from 9 countries: Belgium, Cyprus, Finland, Greece, Italy, Luxemburg, Serbia, Spain and the UK. The project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 952179

For more information

Print friendly pdf

Press contact