01.14 Novel Visualization of Differential TEVAR Success with Koenderink’s Shape Index and Curvedness

K. Khabaz1, S. Dhara3, C. J. Lee4, R. Milner2, L. Pocivavsek2 1University of Chicago,Chicago, ILLINOIS, USA 2University of Chicago Pritzker School of Medicine,Department Of Vascular Surgery,Chicago, ILLINOIS, USA 3University of Chicago,Pritzker School Of Medicine,Chicago, ILLINOIS, USA 4NorthShore University HealthSystem,Department Of Surgery,Chicago, ILLINOIS, USA

Type B aortic dissections (TBADs) have historically been treated with surgery or medical management, depending on the presence of complications. The advent of thoracic endovascular aortic repair (TEVAR) has necessitated a reconsideration of the treatment paradigm due to its reduced peri-operative risk profile but significant long-term complication rate. The guidelines regarding TEVAR versus open surgery still remain in flux and will require greater clarification in the future, particularly with regards to the best choice to reduce complications. We hypothesize that accurate guidance in choosing between endovascular and open repair can be provided by computational characterization of aortic surface geometry, specifically using two indices of surface curvature: the Koenderink shape index, which measures shape, and curvedness, which measures the intensity of the shape. Here, we demonstrate our method of analysis by comparing two TBAD patients with similar non-complicated dissections but distinct post-operative outcomes. 

We study two TBAD patients (A and B) with non-complicated dissections but two distinct post-operative clinical outcomes: aortic remodeling with false lumen thrombosis and failed remodeling with post-procedural complications (type 1A endoleak).  Utilizing CTA imaging, pre- and post-op segmentations of the aortic wall are performed. The curvature tensor is calculated for a mesh of each geometry, allowing us to measure the principle curvatures (k1,k2). Using the Koenderink shape index, s = 2/π arctan((k1+k2)/(k1-k2 )), and curvedness, c = √((k1+k2)^2/2), we measure characterize aortic shape by plotting "2D" histograms of s versus c over each meshed aortic surface. We then calculate statistics of each distribution.

The figure summarizes our results comparing patient A and B. Although the mean shape index, mean curvedness, and variance in shape index values are similar (0.475 vs. 0.481, 0.0352 vs. 0.0412, 0.0640 vs. 0.0641), patient A has a smaller variance in curvedness than patient B (0.00194 vs. 0.000341). This is visually reflected by the larger vertical spread of high intensity regions for patient B. 


Despite relatively similar presentation of non-complicated dissections, the two patients in this study experience different outcomes. Computing the shape index and curvedness values provides a novel and more comprehensive view of the surface geometry. We saw that the variance in curvedness was able to distinguish between a patient with positive post-operative outcomes and a patient with post-operative complications. We hope to determine if our method of analysis holds predictive power with a larger patient set.