SU-C-206-02: Estimating Coronary Artery Plaque Composition with a Combined Dual-Energy and Single-Energy QCT Optimization Model




A study was performed to evaluate the accuracy of a combined dual-energy/single-energy QCT optimization model for estimating the volume fractions of fat, cholesterol, protein, fibrous tissue, and calcium in coronary artery plaque.


This model uses 80kVp-140kVp CT# data to distinguish calcium, and Gaussian component analysis based on the 140kVp CT#'s of the pure materials to distinguish the other constituents. An experimental study was performed with a thorax section phantom (CIRS, Inc.) containing various mixtures of 0.8–3.2mm thick plastic slabs simulating plaque components. The Z(eff), ρ(g/cc) properties of the plaque components and plaque-simulating plastics are: [fat(5.82,0.92), LDPE(5.44, 0.92)], [cholesterol(5.65,1.067), polystyrene(5.62,1.05)], protein(6.73,1.38), Delrin (7.07,1.42)], fibrous(7.51, 1.05), nylon(6.21, 1.14]). Two water-equivalent plastics containing CaCO3 (13% and 22% CaCO3) were also used. Scans of 6.4-mm-thick slabs of each plastic alone within the phantom were acquired to determine the CT#'s of the pure materials. All scans were repeated 3 times under each condition and reconstructed with standard (STD) and BONE kernels on a GE CT750HD CT scanner using a 10-cm reconstruction FOV.


The minimum RMS errors for the various plaque-simulating slab mixtures were 2.2% (BONE kernel) and 4.2% (STD kernel) for a 1.6mm 13% CaCO3, 3.2mm polystyrene, 3.2mm delrin, 3.2mm nylon slab mixture (total=11.2mm). Maximum RMS errors were 7.5% (BONE) and 13.7% (STD) for a mixture containing 1.6mm slabs of 13% CaCO3, LDPE, delrin, and nylon (total=6.4mm).


The accuracy of the method decreased as the thicknesses of the slabs decreased, due to volume averaging. Accuracy was better for the BONE kernel that has higher spatial resolution. Even for this kernel, the errors are substantial for a mixture that is about the size of the coronary artery lumen (2–6mm). Higher spatial resolution and more quantitative reconstruction methods are needed to improve the accuracy of the method.

NIH RO1 HL106545