Automated coronary lumen geometry extraction speeds interpretation
“We know that even in individual coronary arteries there can be multiple different levels of stenoses, multiple different types of lesions,” Rybicki noted. “We also know that low endothelial shear stress (ESS) is the primary local factor for those coronary arteries that form disease.”
Coronary ESS is typically obtained from the lumen geometry derived from intravascular ultrasound (IVUS), although there is increasing interest in non-invasive approaches, Rybicki observed. His research team at Brigham and Women’s sought to demonstrate fully automated 3D coronary lumen geometry extraction based on 320-detector row coronary CT angiography images.
“With this technology, we are able to get a single acquisition of the entire heart over one time point,” Rybicki said.
He reported that contrast-enhanced coronary lumen from an axial 320 x 0.5 mm detector row CT (Toshiba America Medical Systems,) images were automatically segmented using advanced visualization software (Vital Images). The geometric mesh of the segmented lumen surface was over-determined for computational fluid dynamic (CFD) simulation requirements along the circumference. Decimation along the circumference was performed with low-pass spatial filtering. This had the additional advantage of reducing small irregularities in the detected endoluminal surface, Rybicki reported.
Blood velocities through the left main and left anterior descending artery were obtained using commercially available CFD software (CHAM). Rybicki said the research group employed standard assumptions:
• Blood is a Newtonian fluid;
• Viscosity was estimated from hematocrit;
• Zero tangential gradient of velocity at outlet;
• Flow discharging to zero gauge pressure; and
• Uniform velocity profile at the inlet section.
Coronary ESS was calculated as the spatial gradient of the simulated blood velocity field at the inner boundary of the lumen times the viscosity of blood, he reported.
Manual correction of the lumen segmentation was not required, according to the researchers. Axial velocity profiles and ESS maps are obtained from the fully automated 3D coronary lumen geometry extraction. ESS maps demonstrate axisymmetric patterns with circumferential uniformity, Rybicki said.
A significant challenge to CT coronary ESS modeling is the ease at which the lumen geometry can be obtained, Rybicki noted; 320-detector row CT images enable an algorithm to obtain this geometry without user input.
“Non-invasive vascular profiling is certainly in its infancy,” he said. “We’re just now starting to use the technology and we’re starting to apply what we’ve done with IVUS, for quite a long time, to CT.”