Multimodality imaging-driven multifidelity modeling of aortic dissection

点击翻译按钮获取中文摘要

PROJECT SUMMARY. Aortic dissections are responsible for significant morbidity and mortality in young and old individuals alike. Whereas type A (ascending aorta) dissections are treated aggressively via surgery, type B (descending thoracic aorta) dissections are often monitored for long periods to determine the best treatment. These lesions can cease to propagate (i.e., stabilize or heal) or they can propagate further and either turn inward and connect again with the true lumen to form a re-entry tear or turn outward and result in rupture in the case of an compromised adventitia. Notwithstanding the importance of these later events, there is a pressing need to understand better the early processes that initiate the dissection and drive its initial propagation as well as to determine whether the presence of intramural thrombus is protective or not against early or continued propagation. Over the past 5 years our collaborative team has developed numerous new multimodality imaging techniques, biomechanical testing methods, and computational modeling approaches across multiple scales that uniquely positions us to understand better the process of early aortic dissection and the possible roles played by early intramural thrombus development. In this project, we propose to use nine complementary mouse models to gain broad understanding of the bio-chemo-mechanical processes that lead to aortic dissection and to introduce a new machine learning based multifidelity modeling approach to develop predictive probabilistic multiscale models of dissection. These models will be informed, trained, and validated via data obtained from a combination of unique in vitro biomechanical phenotyping experiments (wherein we can, for the first time, quantify the initial delamination process under well-controlled conditions and regional material properties thereafter) and novel multimodality imaging of delamination / dissection both in vitro and in vivo. We will consider, for example, the roles of different elastic lamellar geometries; we will assess separate roles of focal proteolytic activation and pooling of highly negatively charged mucoid material, which can degrade or swell the wall respectively; and we will model and assess the effects of early thrombus deposition within a false lumen. We submit that our new probabilistic paradigm, based on statistical autoregressive schemes and enabled by machine learning tools, could be transformative and lead to a paradigm shift in disease prediction where historical data, animal experiments, and limited clinical input (e.g., multiomics) can be used synergistically for robust prognosis and thus interventional planning. Our work is also expected to lead naturally to an eventual better understanding of the chronic processes associated with dissection via predictive models that are aided by the expected “revolution of resolution” in diagnostic imaging.