SU‐FF‐I‐54: Development and Validation of An Algorithm for Facilitating Observer Studies On the Effect of Radiation Exposure On Diagnostic Accuracy in CT Acquisitions

Abstract

Purpose: In computed tomography (CT), there is increasing concern about detrimental effects of radiation and thus a special need for optimization of image quality and radiation dose. Purpose of this study was to develop and validate software for facilitating observer studies on the effect of radiation exposure on diagnostic accuracy in CT acquisitions taking into account the requested diagnosis. Method and Materials: A low dose CT simulator was developed which adds Gaussian noise to the raw data of the CT scanner. For validation two phantoms were used; a cylindrical water phantom with five different inserts and an anthropomorphic phantom. Scans of both phantoms were performed at different dose levels by changing only the tube current of the acquisition protocol (500‐300‐150‐80‐40‐20 mA). Low dose CT simulations were performed for 500 mA downwards to 20mA in above mentioned steps. Hounsfield Units (HU) and noise (standard deviation of HUs) were measured within six regions of interest (ROIs) in the cylindrical phantom, corresponding to the five different inserts and water; and additionally noise power spectra (NPS) were measured in water. Noise was measured in soft‐tissue equivalent material of the anthropomorphic phantom. Measurements in actually scanned and simulated CT images were compared. Results: The low dose CT simulator yielded similar image quality compared to actual low dose CT acquisitions. Mean differences over all comparisons in noise between actual scanned and simulated images was 5.0±4.2% for the cylindrical phantom and 3.3±2.6% for the anthropomorphic phantom. Evaluation of the NPS showed that spectra are similar with regard to general shape and intensity for the scanned and simulated images. Conclusion: The low dose CT simulator creates images that accurately represent image quality of CT acquisitions at lower dose levels and is suitable to be applied in clinical observer studies.