Arkadiusz Sitek PhD, Presenter: Nothing to Disclose

Aaron Hochberg MD, Abstract Co-Author: Nothing to Disclose

Geoffrey S. Young MD, Abstract Co-Author: Consultant, Pioneer Diagnostics Stockholder, Pioneer Diagnostics Consultant, Thimble Inc Stockholder, Thimble Inc Researcher, Amgen Inc Researcher, AstraZeneca PLC Researcher, Bracco Group Researcher, Celldex Therapeutics, Inc Researcher, F. Hoffmann-La Roche Ltd Researcher, GlaxoSmithKline plc Researcher, Novartis AG Researcher, Siemens AG Research support, Siemens AG Research support, Toshiba Corporation

Kurt Schultz RT, Abstract Co-Author: Employee, Toshiba Corporation

Jeremy Michael Wolfe PhD, Abstract Co-Author: Research Grant, Toshiba Corporation Consultant, The Procter & Gamble Company Consultant, LongShortWay Inc Research Grant, Google Inc

RNS produces simulated reduced-dose images with similar statistical properties to images acquired at the same reduced doses, suggesting that the simulated images may be appropriate for evaluating the consequence of dose-reduction in observer trials.

The reduction of radiation dose in diagnostic CT is a current research priority, but how far can dose be reduced without reducing clinical utility? The ideal evaluation which would require scanning a single patient at multiple doses and comparing the resulting images, is ethically unacceptable. We evaluate an alternative method using software to produce simulated lower doses images by adding noise to full dose raw data. We test if this apporach truly mimics the effects of reduced radiation dose.

No statistically significant differences were detected as the K-S test statistics varied from D=0.026 to 0.064 for different settings of kVp and mAs indicating good agreements of HU histograms obtained from image data acquired on the scanner and that created by the RNS for equivalent doses. No significant differences between 2D FFTs were detected by visual inspection of FFT images.

We compared statistical properties of images from real and simulated scans of the ACR accreditation phantom scanned with 0.5 mm slices with 0.5 mm pitch on a 320 detector CT scanner (Toshiba Aquilion One ®) with four X-ray tube potentials (80, 100, 120, 135 kVp) and 5 tube current settings for each kVp (7 to 330 mAs) which resulted in CTDI ranging from 1.4 to 19.5 mGy. Data were reconstructed with FC41 filter using 0.468 mm pixel size. For each kVp setting, lower-dose data were simulated from raw data corresponding to the highest exposure using the raw noise simulator (RNS). This was done by applying a specific measured noise pattern and calibration table generated for the specific CT system. Images obtained from real and simulated data were compared using HU histograms created by sampling 35,000 pixels (separated from each other) from two uniform regions of the phantom. The histograms were compared using Kolmogorov-Smirnov (K-S) test. In addition, we obtained Fast Fourier Transforms (FFTs) of the slices corresponding to uniform region of the phantom.

Sitek, A, Hochberg, A, Young, G, Schultz, K, Wolfe, J, Evaluation of the Raw Noise Simulator for Radiation Dose-reduction Studies in CT.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12027836.html