Research

Vascular Deformation Mapping is an image analysis technique recently developed in our lab that takes routine gated computed tomography angiography (CTA) data and performs an analysis of 3-dimensional aortic growth in a matter that is more comprehensive and accurate than can be achieved with standard manual diameter-based measurements. We have validated this technique and are currently investigating VDM's uses for better understanding growth patterns, predicting future growth & adverse events and assisting with surgical planning. VDM is being developed for commercial use in conjunction with industry partners at Imbio LLC though the support of an NIH Small Business Innovation Research grant.

4D Flow MRI is and advanced technique that allows for dynamic 3D quantification of blood flow in the aorta and other large vessels and has been applied broadly to study blood flow abnormalities through the body. Our group has specific interest in how the 4D Flow technique can be used to understand blood flow abnormalities in aortic dissection (AD), a disease that leads to two distinct channels for blood flow (true and false lumens), often displays complex anatomy and is associated with poor long-term outcomes in the chronic phase. Blood flow abnormalities in AD are not well understood, but our group has performed preliminary studies showing that there are unique aortic blood flow and pressure abnormalities in the false lumen of patients AD that can be used to understand which patients are at highest risk for aortic growth and need for surgical repair. This work has been supported by prior and current grants from the Radiologic Society of North America (RSNA).

Thoracic aortic aneurysm (TAA) is a common disease; 3 to 23 million patients in the U.S. alone have an enlarged thoracic aorta. This condition is usually asymptomatic and indolent but life-threatening complications (e.g., aortic dissection or rupture) can occur. Patients with TAA undergo regular imaging surveillance, often with CTA, but current imaging surveillance techniques based o aortic diameter measurement are both inaccurate and highly inefficient. In an attempt to improve on current risk assessment techniques in TAA, we have developed a multi-parametric analysis pipeline that takes clinical CTA images and standard clinical information about the patients aortic valve and hemodynamics (extracted from echocardiography and brachial cuff measurements) to build a computational model that allows us to estimate stressed on the aortic wall from the aortic pressure (trasnsmural stress) and the the friction of flowing blood (wall shear stress). These computational maps can be combined with 3D growth data from VDM in an attempt to better understand the mechanisms of TAA growth and improve our ability to predict which patients are at highest risk of progressive growth and other serious complications.

Segmentation and manual measurement of the aorta are time consuming tasks, even for expert readers, but are central components in most aortic analyses. These tasks can be greatly accelerated through image analysis techniques such as deep learning (DL). We are focused on developing and optimizing DL techniques to assist with aortic image analysis though a variety of tasks including automated aortic measurements and registration. The objective of this work is to make aortic image analysis more accurate, efficient and reproducible, and to advance the automation of other algorithms being developed in the lab.