Comparison of Statistical Iterative and Filtered Back Projection Image Reconstruction Techniques for the CT Assessment of Coronary Artery Stents

Purpose
Statistical iterative reconstruction in image space (IRIS) may allow more accurate delineation of anatomic structures and decrease beam hardening artifacts and image noise compared to traditional filtered back-projection (FBP) image reconstruction techniques. The aim of this study was to prospectively compare FBP and IRIS image reconstruction techniques for the CT evaluation of coronary artery stents.

Method and materials
Coronary CT (Definition Flash, Siemens) angiography raw data of 14 patients with coronary stents were reconstructed with FBP and IRIS, using dedicated stent algorithms (“B46” and “I46”, respectively). Image quality was subjectively rated by two independent observers. The observers manually measured luminal diameters at 5 different points (proximal to the stent, proximal end of the stent, mid-portion, distal end, and directly distal to the stent) using curved multiplanar reformations. Image noise was measured as the HU standard deviation in the aortic root. Stent volumes were measured with dedicated volume software which included only voxels with an attenuation of 20% above the maximum intraluminal contrast media attenuation. Variables were compared with chi-square statistics for categorical variables and with Student’s unpaired t test for continuous variables. Interobserver agreement was calculated using k-statistics.

Results
Image quality of IRIS studies was rated significantly (p= 0.02) higher than that of FBP reconstructions. Mean manual luminal diameter measurements were 0.20±0.77mm with FBP and 0.21±0.68mm with IRIS (p=0.76). Interobserver agreement between reader 1 and 2 was higher with IRIS than with FBP (k=0.954 versus k=0.800). Image noise was significantly lower with IRIS (23±12) than with FBP (45±17, p=0.03). Stent volumes measured significantly lower in studies that were reconstructed with IRIS compared to those that were reconstructed with FBP (0.16±0.4cm3 versus 0.19±0.6 cm3, p=0.04).

Conclusion
IRIS significantly reduces image noise and improves the reproducibility of CT coronary stent evaluation compared with FBP. The lower stent volume measurements indicate a significant reduction of beam-hardening (“blooming”) artifacts with IRIS, enabling more accurate stent delineation.

Clinical relevance/application
Iterative image reconstruction techniques incrementally improve the assessment of coronary artery stents by reducing image noise and beam-hardening artifacts.

Submission Type: Scientific Presentations RSNA 2010

Authors: A Ramachandra, BSC, Charleston, SC; U J Schoepf, MD; M Weininger, MD; S Tipnis; W Huda, PhD;
T Henzler, MD; et al.