The aims of WP2 are:
– To identify population subgroups & image acquisition key-factors for clinical data categorization, using information collected from screening equipment manufacturers and clinical stakeholders;
– To categorise clinical data into subsets for successive AI tool training and benchmarking; to assess data quality, identifying the subsets where data are lacking;
– To supplement lacking subsets by generating synthetic data with physics-informed models able to replicate breast imaging pipeline, ranging from digital breast phantom construction and compression, up to synthetic mammogram production;
– To extend the datasets used for AI tool benchmarking by using data perturbation and data augmentation techniques based on statistical and ML-based approaches.