Ehsan Samei PhD, Presenter: Research Grant, Siemens AG Research Grant, General Electric Company Research Grant, Carestream Health, Inc Consultant, KUB Technologies, Inc

Grant M. Stevens PhD, Abstract Co-Author: Employee, General Electric Company

James G. Colsher PhD, Abstract Co-Author: Employee, General Electric Company

Donald P. Frush MD, Abstract Co-Author: Nothing to Disclose

To design and evaluate a CT noise simulation tool that enables virtual dose reduction of clinical CT images with realistic noise magnitude and texture. The tool enables clinical trials of iterative reconstructions using images with realistic subtle lesions and without the need for repeated exposures.

The simulation tool operates on CT projection data. Starting from the mAs of the acquired image data, the system determines the level of increased noise in the projections for a targeted reduced mAs value. Noise is added to the projection data taking into account the scanner characteristics (eg, electronic noise) and patient properties. Lesion masks based on real lesion specifications are similarly added to the projection data. The tool was evaluated in terms of noise magnitude and texture, CT number, and resolution on a commercial CT scanner (GE CT750 HD) across a range of kVps (80-140) and mAs (50-200) with analytical (FBP) and iterative (MBIR) reconstructions. 692 actual and simulated dose-reduced clinical images were also compared in appearance using a 4AFC experiment with 8 observers.

There was less than 10% difference in noise magnitude between real and simulated images. Resolution and noise texture were identical between the two datasets. CT numbers were accurate to within 1.5 HU. Observer performance indicated that the simulated images with FBP were virtually identical to the real ones (22% ± 3% chance identification). The images were more distinguishable with MBIR (16% ± 5%). However, the simulated images were judged to be essentially identical to original images in terms of diagnostic quality. Lesions appeared realistic in the simulated images.

Justified reduction of dose is paramount in today’s practice of radiology. Such reduction should ideally rely on clinical images with subtle pathology with serial predictable reduced dose levels. However, the former is rare to come by and the later is impractical due to ethical consideration of exposing a patient, especially children, to radiation multiple times. The tool introduced in this study addresses both of these limitations. The tool affords a unique benefit in the evaluation of the dose reduction potential of iterative reconstructions.

The noise and lesion simulation program provides a key technology to enable systematic study of dose reduction and its impact on diagnostic quality.

Samei, E, Stevens, G, Colsher, J, Frush, D, Design and Evaluation of a Projection-based Noise Program for Dose Reduction Studies for CT.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12028903.html