MagIC researchers have made significant contributions to the area of Health over the last five years, particularly through the BINDER FCT project coordinated by Leonardo Vanneschi in collaboration with the Champalimaud Foundation. Using novel and cutting-edge machine learning and deep learning methods, this project unveiled previously unknown distinctions between breast and axilla tumors, examining diverse types of data, including radiomic features.
The study found that peritumoral radiomic features play a pivotal role in breast tumors, while the internal tumor area is significant in axilla tumors, suggesting higher predictability for breast tumors compared to axilla tumors. However, leveraging advanced regularization techniques, our models achieved an unprecedented 87% accuracy in prognosticating complete pathologic responses of axilla tumors, a groundbreaking result in the literature for this tumor type. Additionally, within the BINDER project, a deep learning framework was devised for tumor cell detection and quantification from microscopic images, alongside pioneering strategies for prostate segmentation.
The BINDER project's empirical findings offer insights into previously unknown distinctions between breast and axilla tumors, with significant implications for predicting complete pathologic responses of axilla tumors, advancing the field of Health through cutting-edge machine learning and deep learning methods.
Project: BINDER: Improving Bio-Inspired Deep Learning for Radiomics. Funded by the FCT (Fundação para a Ciência e a Tecnologia), Portugal, FCT, PTDC/CCI-INF/29168/2017. Period: 2018–2021. Coordination: NOVA IMS. Partners: NOVA IMS; Faculty of Science of the University of Lisbon; University of Coimbra; Champalimaud Foundation.