• David B Larson, Lily L Wang, Daniel J Podberesky, and Marilyn J Goske.
• Department of Radiology, Cincinnati Children's Hospital Medical Center, 3333 Burnet Ave, MLC 5031, Cincinnati, OH 45229.
• Radiology. 2013 Oct 1; 269 (1): 167-76.

PurposeTo develop and validate a mathematical radiation dose optimization model for computed tomography (CT) of the chest, abdomen, and pelvis.Materials And MethodsThis quality improvement project was determined not to constitute human subject research. A model for measuring water-equivalent diameter (DW) based on the topogram was developed and validated on each axial section in eight CT examinations of the chest, abdomen, and pelvis (500 images). A model for estimating image noise and size-specific dose estimates (SSDEs) using image and metadata was developed and validated in 16 examinations of anthropomorphic phantoms. A model to quantify radiologist image quality preferences was developed and applied to evaluations of 32 CT examinations of the abdomen and pelvis by 10 radiologists. The scanners' dose modulation algorithms were modeled and incorporated into an application capable of prediction of image noise and SSDE over a range of patient sizes. With use of the application, protocol techniques were recommended to achieve specific image noise targets. Comparisons were evaluated by using two-tailed nonpaired and paired t tests.ResultsThe mean difference between topogram- and axial-based DW estimates was -3.5% ± 2.2 (standard deviation). The mean difference between estimated and measured image noise and volume CT dose index on the anthropomorphic phantoms was -6.9% ± 5.5 and 0.8% ± 1.8, respectively. A three-dimensional radiologist image quality preference model was developed. For the prediction model validation studies, mean differences between predicted and actual effective tube current-time product, SSDE, and estimated image noise were -0.9% ± 9.3, -1.8% ± 10.6, and -0.5% ± 4.4, respectively.ConclusionCT image quality and radiation dose can be mathematically predicted and optimized on the basis of patient size and radiologist-specific image noise target curves.© RSNA, 2013.

19 keywords...

hide…