RSNA 2014 

Abstract Archives of the RSNA, 2014


A Four-dimensional Segmentation Algorithm for Accurate Reproduction of 3D-printed Coronary Arteries from Clinical Coronary CT Angiography

Scientific Papers

Presented on December 2, 2014
Presented as part of SSG07: Informatics (3D, Quantitative and Advanced Visualization)


Dimitris Mitsouras PhD, Presenter: Nothing to Disclose
Kurt Schultz RT, Abstract Co-Author: Employee, Toshiba Corporation
Frank John Rybicki MD, PhD, Abstract Co-Author: Research Grant, Toshiba Corporation
Gerald Thomas Grant DMD, MS, Abstract Co-Author: Nothing to Disclose


To reproduce accurate patient-specific coronary tree phantoms from clinical CTA and use them for in vitro CTA experiments to elucidate the intraluminal kinetics of iodinated contrast agent giving rise to the Transluminal Attenuation Gradient (TAG).


A “four-dimensional” segmentation algorithm was developed taking advantage of cardiac motion in the CTA cardiac phase window. The algorithm iteratively combines commercial segmentations (Vitrea 6.2) of the coronary tree in individual phases to a single segmentation. This is accomplished by iteratively solving for motion between cardiac phases using least squares optimization and refining the “corresponding point” problem at each iteration. The corresponding points of individual segmentations are then combined into a single segmentation of the lumen plus its motion in-between cardiac phases. The algorithm was applied to 3 phases (70,75,80% of R-R) of a coronary CTA and the resulting endoluminal surface, extended with Luer connectors, was directly output in STL file format and printed on a Viper 2SI stereolithography machine (courtesy, Gerald T. Grant and Shayne Kondor, 3D Medical Applications Center, Walter Reed National Military Medical Center). The printed phantom was attached to a dual injector (Bracco Diagnostics) with one syringe loaded with saline and the other a mixture of 10:90% contrast:saline and scanned at 0.5sec intervals using a volume CTA protocol on 320-detector row CT (Toshiba) from a starting state of saline and during an injection protocol of 10s@0.1ml/s of mixture, followed by 6s@0.5ml/s of mixture, and finally 6s@1ml/s saline. 


CTA of the printed phantom during contrast passage resulted in contrast opacification curves along the length of the coronary (Figure) of similar TAG magnitude as seen in vivo, between -26 HU/cm (upslope) to 9.5 HU/cm (downslope). Turbulence in the fluid jet due to an incidental luminal stenosis was readily detected and quantified by the CT Hounsfield units during contrast arrival.


The effect of blood flow on CTA contrast enhancement is sufficiently large to be detected by current CT technology and gives rise to the TAG observed in vivo


High-fidelity 3D-printed coronary artery phantoms reproduced from clinical CTA will enable understanding the principles underlying TAG measurements and how they may be properly used to detect functionally significant coronary artery disease.

Cite This Abstract

Mitsouras, D, Schultz, K, Rybicki, F, Grant, G, A Four-dimensional Segmentation Algorithm for Accurate Reproduction of 3D-printed Coronary Arteries from Clinical Coronary CT Angiography.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL.