An explainable AI system for automated COVID-19 assessment and lesion categorization from CT-scans

Highlights：

• A novel lung-lobe segmentation network outperforming state of the art models;

• The segmentation network drives a classification network in first identifying CT scans of COVID-19 patients, and, afterwards, in automatically categorizing specific lesions;

• An interpretation of the decisions made by the employed models and discover that those models focus on specific COVID-19 lesions for distinguishing whether a CT scan pertains COVID-19 patients or not;

• The whole AI pipeline is integrated into a web platform to ease use for radiologists, supporting them in their investigation on COVID-19