By the Fundacao D. Anna Sommer Champalimaud e Dr. Carloss Montez Champalimaud (CF)
Champalimaud Foundation (CF) is a private non-profit Portuguese institution, inaugurated in 2010 with the aim of advancing basic and translational world-leading research in neurosciences, physiology, and cancer, in close collaboration with tailored clinical care. Within the ProCAncer-I project, CF will act as both clinical and technical partner. In total, it is expected that CF will contribute with about 1200 multi-parametric Magnetic Resonance Imaging (mpMRI) exams. From a clinical provider’s perspective, there will be an involvement from the Imaging, Urology and Histopathology CF Departments in the data collection, image annotation and validation of AI models. These departments have shown extensive experience in diagnosing, treating, and monitoring PCa patients throughout the disease continuum. On a complementary fashion, the Computational Clinical Imaging Group (CCIG) at CF, led by Dr. Nickolas Papanikolaou, the Scientific Coordinator of ProCAncer-I, will focus on relevant technical aspects of the project. One of the main responsibilities of CF as a ProCAncer-I partner is the coordination of WP5 on the development of the master models, based on the retrospective heterogeneous data from multiple vendors and institutions. This comprises the supervision of tasks such as the retrospective data upload and annotation, development of pre-processing pipelines – that address distortions (bias field) correction, motion-related artifacts, noise reduction, and data harmonization –, and the development and evaluation of AI master models. Another task with CF involvement is the creation of vendor-neutral models trough federated-learning strategies. Additionally, the clinical use case four on the biological validation of AI models through a radiologic-histopathologic correlation for side-by-side comparison of both image formats, will be performed at CF using whole-mount 3D-printed prostate molds from a subpopulation of radical prostatectomy patients. Finally, all models are to be compared from a radiomics and deep learning modeling perspective. During the last 5 years, the CCIG group has been involved in EU-funded projects and developed an international radiomics network with more than 15 partners coming from Europe, US, and Brazil. The purpose of the CCIG investigations is to apply AI and radiomics algorithms to solve problems in prostate, pancreas, rectal, lung, and breast cancer, gliomas, and multiple myeloma, among others. In conclusion, with the previously described set of expertise, in combination with a lot of work and motivation, it is intended to deliver robust, fair, and interpretable AI models, for addressing prostate cancer open questions, and to deploy them into clinical practice.