RSNA 2005 

Abstract Archives of the RSNA, 2005


SST14-07

Improving the Contrast and Noise Characteristics of Simulated Mammograms

Scientific Papers

Presented on December 2, 2005
Presented as part of SST14: Physics (Diagnostic Image Quality)

Participants

Michael Ronald Chinander PhD, Presenter: Nothing to Disclose
Robert Mark Nishikawa PhD, Abstract Co-Author: Nothing to Disclose

PURPOSE

We have previously developed a model of a radiographic phosphor screen in order to create simulated mammograms. In this study we investigated the addition of scatter and film noise components to improve the accuracy of the model.

METHOD AND MATERIALS

A monte carlo-based scatter point spread function (PSF) available in the literature was convolved with the input x-ray distribution images to determine that amount of scatter. We used a PSF which corresponded to a 4 cm thick breast and 5 cm diameter field of view. The resulting scatter to primary ratio (SPR) and change in contrast was used to the evaluate the effect of adding scatter and compared to expected values. Film grain noise was added to the simulation output images by generating zero-mean gaussian white noise with the variance of the noise dependent upon the local optical density. The film grain noise was filtered with the noise transfer function of the film prior to adding it to the output image. A high resolution, low noise radiographic image of cadaver breast tissue was also used as input to the simulation and compared to a radiograph of the cadaver breast using a conventional mammographic screen film system.

RESULTS

The SPR of the simulation agreed well with the expected value, within 6% of the expected SPR. The addition of scatter resulted in a reduction in contrast by 22% as expected by theory. The standard deviation of a region from the simulated mammogram was within 11% of the standard deviation of the corresponding region from the screen film mammogram.

CONCLUSION

The addition of scattered radiation and film grain noise improves our model to produce simulated the mammograms. The simulated mammograms more closely match the noise and contrast characteristics of real mammograms.

Cite This Abstract

Chinander, M, Nishikawa, R, Improving the Contrast and Noise Characteristics of Simulated Mammograms.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4413236.html