Effect of dose metrics and radiation risk models when optimizing CT x-ray tube voltage

We investigated the effect of different CT dose metrics, as well as the implications of various radiation risk models, on the optimization of x-ray tube voltage (kV) in CT. Soft tissue attenuation characteristics and noise levels, obtained from CT scans of a Rando phantom, were used to compute contrast-to-noise ratios (CNR) at x-ray tube voltages between 80 and 140 kV. Four CT dose metrics were evaluated: (a) CTDIair, (b) weighted CTDIw, (c) organ dose (Dorgan), and (d) effective dose (E). All doses were obtained using the ImPACT CT Dosimetry software package. Soft tissue CNR was adjusted by the modification of the mAs by assuming that CNR(2) was proportional to mAs. Optimization criteria were: (a) maintaining a constant CNR at each kV and identifying the value that minimizes patient dose; and (b) maintaining a constant dose at each kV and identifying the value that maximizes CNR. We also investigated the implication for optimization strategies assuming that radiation risk is proportional to En, with n varying between 0 and 2. Optimizing with respect to phantom measurements (i.e., CTDIair and CTDIw) could generate results that differed quantitatively and qualitatively from those obtained using patient doses (i.e., Dorgan and E). For head CT scans, 140 kV offered the lowest patient doses as well as the highest CNR, whereas in abdominal scans 80 kV was optimal. Use of an optimal kV for CT imaging over current practice of using 120 kV might reduce patient doses by 10-15%, or improve CNR by 5-10%. Assuming that the risk was proportional to En made no difference to the optimal kV for positive values of n up to 2. We conclude that (a) CT optimization with respect to kV should generally be performed with respect to the patient effective dose, (b) neither CTDIair nor the body CTDIw are appropriate for use in CT optimization, (c) the range of current radiation risk models should not affect the optimal kV value in CT imaging.