RSNA 2011 

Abstract Archives of the RSNA, 2011


A Comparison of CT Dose Reduction Methods with Respect to Breast Dose and a Task-based Image Quality Metric

Scientific Informal (Poster) Presentations

Presented on December 1, 2011
Presented as part of LL-PHS-TH: Physics


Franco Rupcich BS, BEng, Presenter: Nothing to Disclose
Andreu S. Badal PhD, Abstract Co-Author: Nothing to Disclose
Iacovos Kyprianou PhD, Abstract Co-Author: Nothing to Disclose
Taly Gilat Schmidt PhD, Abstract Co-Author: Nothing to Disclose


To compare CT dose reduction methods by quantifying breast dose and task-based image quality.


A signal-known-exactly, background-known-exactly signal-to-noise ratio (SNR) metric quantified image quality as it considers contrast, spatial resolution, and noise and its correlations, which may vary across dose reduction methods. The task was a 5mm diameter iodine sphere in the center of the heart of a voxelized female phantom. Noise-free projection data was obtained via ray-tracing, and images were reconstructed by filtered backprojection. The noise-free image of the phantom without the sphere was subtracted from the noise-free image with the sphere, resulting in the signal image. One thousand independent noisy images were obtained by adding Poisson noise to the projections before reconstructing each image. The covariance matrix was estimated by extracting an ROI around the sphere from each noisy image. To reduce the noise of the estimate of the covariance matrix and to make its inversion possible, the off-diagonal elements were assumed to have the same correlation scaled by the location specific variance. The task-based SNR was calculated from the covariance matrix and signal image. A database of dose to different organs as a function of photon energy and projection angle was previously estimated using Monte Carlo techniques and validated for the voxelized phantom. The database was used to estimate dose to the breast for each scan protocol studied. The SNR2 per breast dose was compared for a full scan and a partial scan, which has been proposed to reduce dose to the breast.


For a 120 kVp cardiac CT scan when the dose to the breast is maintained constant, the SNR2 of the partial scan technique (232° centered about PA) is 1.3 times higher than the full scan technique.


Results suggest partial scan techniques may be beneficial in reducing dose while maintaining image quality. We quantified the tradeoff between image quality and breast dose using a spatial, task-based SNR metric normalized by dose-to-breast. Future work will use this technique to compare tube current modulation, reduced kVp, and breast shielding in terms of SNR2/dose.


This work compares CT dose reduction methods with respect to dose-to-breast concurrently with task-based image quality.

Cite This Abstract

Rupcich, F, Badal, A, Kyprianou, I, Schmidt, T, A Comparison of CT Dose Reduction Methods with Respect to Breast Dose and a Task-based Image Quality Metric.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL.