Pulmonary embolism (the blockage of a pulmonary artery by a blood clot) is the third cause of cardiovascular death in Europe. Upon diagnosis confirmation, clinicians evaluate the patient prognosis based on risk stratification models. The management of patient, thus its outcome, highly depend on this risk stratification. Recent studies confirmed that patient stratification based on computed tomography pulmonary angiogram (CTPA) are not well correlated to patient prognosis. As more than 90% of PE are diagnosed with CTPA, it is crucial to improve the ability of CTPA to stratify the prognosis of patients with PE in order to improve their management. The goal of this project is to develop better risk stratification models from the routinely performed CTPA. To build these models, we will propose new morphological PE biomarkers from CTPA. We aim at providing the first open software to automatically compute established and our proposed PE biomarkers from the routinely performed CTPA exam. This software will help clinicians to drastically reduce the time management of PE patients and thus improving their outcome.

Through this ambitious clinical objective, this project will also tackle two more fundamental questions:  1/ the study of new vascular alteration PE biomarkers based on physiological hypotheses. Finding such biomarkers is likely to open clinical research questions, and ultimately  lead to new guidelines for PE patient management. 2/ the development of a robust and generalizable vascular segmentation algorithm, based on a new 3D deep learning-based tracking and ensuring its connectivity, along with the modelling of the vascular network. The interest of such algorithms goes beyond the clinical question of this project, as vascular modelling is the first crucial step of any computed-assisted diagnosis or treatment of cardiovascular applications.