Transforming cancer care with AI and Big Data
Despite technological advances, early cancer detection, accurate tumour classification and treatment monitoring remain challenging. Medical imaging provides important information about the location, size and metabolism of a tumour, benefitting diagnosis and treatment planning. Novel tools based on explainable AI and machine learning, if trained on representative and fair health data, can improve the accuracy of medical imaging and significantly support the clinical work flow in various ways.
AI-powered toolbox
The key objective of the EU-funded INCISIVE project was to leverage the power of big data and AI to enhance cancer imaging. The project focuses on four prevalent cancers: namely breast, colorectal, lung and prostate. “We have developed an AI-powered toolbox for healthcare professionals and a reusable federated data repository of millions of medical images and clinical data,” explains project coordinator Gianna Tsakou. The INCISIVE AI toolbox offers explainable decision-support services for healthcare professionals, covering various imaging modalities such as CT, PET-CT, MRI, mammography, ultrasound, X-ray and histopathological images. It features and integrates 28 innovative AI models trained and validated on millions of cancer images and clinical data, predominantly sourced from the INCISIVE federated data repository. The toolbox provides decision-support services such as abnormalities classification, patient prioritisation, lesion segmentation and localisation, cancer diagnosis and staging, as well as risk of metastasis prediction. Additional features include visualisation through augmented reality and image-to-report transformation that enhance usability and make it easier for professionals to integrate AI insights into clinical workflows.
Data repository
Despite the potential of health data to fuel AI innovations, existing datasets are scattered and lack interoperability, while ethical data sharing is highly challenging. To address this data fragmentation, the INCISIVE project developed a pan-European federated data repository. This repository houses over 5.5 million anonymised cancer images and accompanying clinical data from more than 11 000 individuals across 14 clinical centres. “The repository’s federated architecture and ethical data sharing mechanisms ensure high standards of data privacy and security, allowing data holders to maintain control over their data while enabling AI training, validation and research,” emphasises Tsakou. The INCISIVE platform provides a secure and controlled environment for data discovery and ethical processing, even if the data is distributed in different clinical settings and EU countries. It integrates all project services and functionalities, offering, among others, tools for data anonymisation, collection, curation and annotation, supporting GDPR-compliant data sharing.
INCISIVE impact and future directions
The increasing amount and availability of cancer imaging and other health data, in combination with advances in AI and machine learning, provide unprecedented opportunities for improving cancer care and management. By taking all appropriate ethical and legal measures to ensure health data privacy and security, the INCISIVE cancer data repository will facilitate future cancer-related research significantly contributing to the European Health Data Space. The next steps involve enhancing interoperability with other health data repositories and research infrastructures, through the Cancer Image Europe initiative. By fostering collaboration and supporting the integration of AI services into clinical practice, INCISIVE aims to sustain its impact and drive future innovations in cancer care.
Keywords
INCISIVE, AI, toolbox, data repository, cancer imaging, machine learning