| Project Detail |
Artificial Intelligence for Breast cancer diagnosis
Breast cancer poses a significant threat to many EU citizens. Despite European efforts to mitigate this threat through early and accurate diagnosis, the nascent stage of most precision medicine means that patients often lack precise treatment following neoadjuvant chemotherapy (NAC). With the support of the Marie Sklodowska-Curie Actions programme, the GRANITE project aims to leverage the potential of AI for digital diagnostics and enhance its capabilities. To achieve this, the project will address current challenges faced by AI technology in diagnosis. It will equip AI with state-of-the-art image analysis, healthcare data, and pre-train multiple deep learning (DL) models. Subsequently, the project will deploy these pre-trained DL models to demonstrate, fine-tune, and validate this solution. Through these efforts, GRANITE seeks to advance the field of digital diagnostics for breast cancer and improve patient outcomes.
In EU-27, it is estimated that >355,000 were diagnosed with breast cancer (BC) in 2020. Initiatives to reduce this burden in Europe involve early and precise diagnosis in the standard-of-care management to decrease unnecessary or insufficient treatment. Notably, precision medicine in BC is still in its infancy and is becoming even more critical with neoadjuvant chemotherapy (NAC), a standard of care treatment protocol in HER2-positive (HER2+) and triple negative (TN) subtypes. Recently, the use of digital diagnostics with AI is gaining momentum since it shows great promise towards accelerating personalized BC patients’ pre- and post-NAC predictions. While AI paves the way to next generation diagnostics, this has yet from been translated as a) it mostly operates in single data modalities that fail to capture the complex disease alterations, b) integrated AI usually suffers from data incompleteness usually leading to models trained with limited data that fail to generalize to new patients and/or that are not able to integrate partially observed multimodal information from the whole population. GRANITE focuses to address these unmet needs and goes beyond the state of the art, fusing the most relevant standard of care data (radiology, pathology, clinical, demographic), leveraging novel AI and state-of-the-art image analysis (radiomics) based on the extensive previous work of the experienced researcher (ER). GRANITE will deploy pre-trained DL models from the ER and fine-tuned, technically validated and clinically evaluated using more than 150 non-metastatic BC cases from the Bank of Cyprus Oncology Centre (BoCOC). ER’s participation in the AI4HI project will help transcend FUTURE-AI guidelines (Fairness, Universality, Traceability, Usability, Robustness and Explainability; future-ai.eu) into GRANITE to generate real-world evidence and make our AI clinically sound, ethically aware and technically applicable, and to promote AI trust and and acceptance in BC management. |
