Optimization of CT calcium scoring doses on the General Electric Discovery single-photon emission computed tomography/CT D670, 8-slice scanner.

Manufacturer-recommended exposure mA was typically resulting in 3-5 times greater patient doses for calcium score scans compared with other dedicated CT scanners at Nottingham University Hospitals. Image noise was used as a measure of image quality in phantom and patient data. The noise was quantified from the standard deviation in Hounsfield units within regions of interest in the myocardium. Noise in phantom data was found to vary linearly with the inverse square root of the applied mAs. It was assumed that a linear relationship would also apply to patient data but it was predicted that the linear gradient would vary between patients owing to differing patient size and composition. This noise model was used to calculate the exposure mA required to achieve a target noise level of 25 Hounsfield units in the myocardium for each patient. To maintain the image quality for patients of different sizes, three measures of size, weight, body mass index (BMI) and lateral dimension, were all tested for goodness of fit to the noise model. It was found that BMI correlated best with the noise model for small patients, and therefore, BMI was chosen as a measure of patient size for the revised mA table. Using this methodology, doses to small patients were reduced by a factor of four compared with manufacturer-recommended settings.