Mastering the vast world of PRE-ACT data

High quality data is the key to pragmatic and successful development of Artificial Intelligence (AI) in the clinic to support personalized medicine. Maastro Clinic (Netherlands), its spin-off Medical Data Works, and the French company TheraPanacea are supporting the PRE-ACT consortium in the collection, harmonization, and standardization of multi-modal (images, electronic health records, genomics) and multi-centric data from our institutions. They are working together to build a database including 6000+ patients with breast cancer recruited in REQUITE, CANTO and HYPOG-01 observational and clinical trials. The database includes a comprehensive set of potential clinical, dose-volume, genetic and radiomic predictors which will be used to estimate the risk of radiotherapy-induced toxicities (i.e., arm lymphedema, telangiectasia, skin induration, skin hyperpigmentation, erythema, and edema). We have been through complex yet fascinating challenges while extracting and integrating the dose-volume measures from radiotherapy plans which have been collected in extremely diverse formats. Safe de-identification process of the genomic data and appropriate data augmentation techniques on missing records will remain to be tackled in next steps.