Two-Dimensional Quantitative Coronary Angiographic Models for Bifurcation Segmental Analysis: In Vitro Validation of CAAS Against Precision Manufactured Plexiglas Phantoms

Background: Quantitative coronary angiography (QCA) analysis for bifurcation lesions needs to be standardized. Objectives: In vitro validation of two models for bifurcation QCA segmental analysis. Methods: In the latest edition of the Cardiovascular angiography analysis system (CAAS 5v8, Pie Medical Imaging, Maastricht, The Netherlands) a 6-segment model for two-dimensional coronary bifurcation analysis was implemented next to the currently available 11-segment model. Both models were validated against 6 precision manufactured plexiglas phantoms, each of them mimicking a vessel with three successive bifurcation lesions with variable anatomy and Medina class. The phantoms were filled with 100% contrast agent and imaged with a biplane gantry. Images acquired in anteroposterior (AP) direction by either C-arm and at 300 right and left anterior oblique angulation were analyzed by two independent analysts, blinded to the actual dimensions. Manual correction of the contours was not allowed. Measurements for minimal lumen diameter (MLD), reference vessel diameter (RVD), percent diameter stenosis (DS) and bifurcation angle (BA) were compared with the true phantom dimensions. Results: In AP views the accuracy and precision (mean difference +/- SD) of 11- and 6-segment model for MLD, RVD, and DS were 0.065 +/- 0.128 mm vs. 0.058 +/- 0.142 mm, -0.021 +/- 0.032 mm vs. -0.022 +/- 0.030 mm, and -2.45% +/- 5.07% vs. -2.28% +/- 5.29%, respectively. Phantom MLD values <= 0.7 mm were systematically overestimated; if excluded, MLD accuracy and precision became 0.015 +/- 0.106 mm and 0.004 +/- 0.125 mm for the 11- and 6-segment model, respectively. Accuracy and precision for BA were -2.2 degrees +/- 3.3 degrees. Interobserver variability for MLD, RVD, DS, and BA for either model was <= 0.049 mm, <= 0.056 mm, <= 2.77%, and 1.6 degrees, respectively. Agreement between models for MLD, RVD, and DS was +/-0.079 mm, +/-0.011 mm, and +/-2.07%. Accuracy and precision for diameter-derived parameters were slightly decreased in angulated projections; precision for BA measurements dropped to 6.1 degrees. Conclusions: The results of both models are highly reproducible and for phantom MLD values >0.7mm in excellent agreement with the true dimensions. (C) 2011 Wiley-Liss, Inc.