Peak SNR in automated coronary calcium scoring: selecting CT scan parameters and statistically defined scoring thresholds.

PURPOSE: Development and verification of peak signal-to-noise ratio (SNR(P)) equations for determining optimum CT scanning and scoring parameters for a new automated coronary calcium scoring program (N-vivo). Experimental evaluation of the new program for scoring small calcium hydroxyapatite (CaHA) microspheres with small voxel CT images.

METHODS: Theoretical SNR(P) equations were developed using measures of noise, resolution, contrast, scatter, and x-ray photon energy. A coronary calcium scoring test phantom containing very small CaHA microspheres was scanned simultaneously with a calibration phantom at three kVps, three voxel sizes, and three phantom sizes. Agatston and calibrated mass scores, noise standard deviations, peak noise, and peak signal voxel intensities were measured by the N-vivo method for individual microspheres and in patient CT scans.

RESULTS: The SNR(P) equation was predictive of the optimum voxel size, kVp, and phantom size, and allowed automated computation of mass scoring thresholds specific to each patient and CT scan. The smallest microcalcification scored in the full sized phantom with the N-vivo method by calibrated mass score (volume 0.075 mm3, mass 0.17 mg) was approximately four times smaller than that scored with the Agatston method (0.27 mm3, mass 0.63 mg).

CONCLUSIONS: The SNR(P) equation can be used to model and optimize calcium scoring and CT scan parameters. The common assumptions that noise is too high in thin slice CT or requires high radiation dose for CAC scoring are shown to be misleading. The N-vivo method showed higher detection sensitivity for small microspheres and more consistent mass scores at different beam energies than the conventional Agatston method. Advanced calcium scoring methods with higher sensitivity may allow improved monitoring of plaque changes and provide earlier detection of atherosclerosis.