Ruben Edward van Engen, Presenter: Nothing to Disclose

Ramona Bouwman, Abstract Co-Author: Nothing to Disclose

Wouter J. H. Veldkamp PhD, Abstract Co-Author: Nothing to Disclose

Adriana MJ Bluekens MD, Abstract Co-Author: Nothing to Disclose

Mireille Broeders PhD, Abstract Co-Author: Nothing to Disclose

Gerard J. den Heeten MD, PhD, Abstract Co-Author: Research Consultant, Koninklijke Philips Electronics NV Stockholder, Sigmascreening

To investigate the relationship between anatomical information in mammography images and dose level using power spectra analysis and to study the potential of additional low dose views in screening mammography.

Images of 5 amputated breasts have been made on a Hologic Selenia at different dose levels (10 to 100% of AEC dose). The X-ray spectrum was chosen by the AEC. Power spectra of all images were calculated. A Power law function according to Burgess was fitted at low spatial frequencies where anatomical information is dominant and a linear fit at high spatial frequencies where system noise is dominant. System noise is assumed to be approximately white. Subsequently the intercept of this power law function and the system noise is determined, which was chosen as a figure of merit to assess the highest frequency at which anatomical information is dominant.

The intercept between both fits shifted towards higher frequencies at higher detector dose levels. At AEC dose the intercept was in the range of 3.0 to 4.0 lp/mm and at 10% of the AEC dose in the range of 0.6 to 0.7 lp/mm, corresponding with object sizes in the range of respectively 125 to 170 micron and 600 to 700 micrometer.

The spatial frequency of the intercept between the system noise fit and power law fit decreases with dose. At AEC dose the smallest objects size at which the anatomical information is dominant was approximately 125 micrometer. At 10% of the AEC dose anatomical information of larger structures (> 700 micron) is dominant in the images. A low dose image in an additional view might be sufficient to determine whether a large structure is an projection of overlying tissue or real. Techniques to reduce high frequency noise might also be used to improve the quality of the low dose image without decreasing anatomical information.

Additional low dose images contain information about larger structures, which might aid the radiologist’s interpretation without increasing the dose to the patient substantially.

van Engen, R, Bouwman, R, Veldkamp, W, Bluekens, A, Broeders, M, den Heeten, G, Information Content of Mammographic Images at Different Dose Levels.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12043880.html