Steven Raeymaeckers, Presenter: Nothing to Disclose

Nico Buls, Abstract Co-Author: Nothing to Disclose

Henri Souchay, Abstract Co-Author: Nothing to Disclose

Johan De Mey, Abstract Co-Author: Nothing to Disclose

Catherine Breucq, Abstract Co-Author: Nothing to Disclose

To construct a breast phantom for image quality evaluation in digital breast tomosynthesis.

A multilayer phantom with a diametre of 13 cm was constructed from a base of liquid polyurethane prepolymer rubber and contains two layers of masses, one with microcalcifications and one with a structured noise pattern. Within the two first molds we simulated masses of different diameters (range 3,9 – 18,1 mm), by introducing circular spheres from a type of liquid rubber with a 4% higher specific gravity. Attenuation properties of both rubbers were determined experimentally with a MoMo 28kV photon spectrum. The third mold contains clusters of pulverised eggshell (range 0,1 - 0,3 mm) to simulate microcalcifications. The last mold contains two pieces of preserved cow-udder, to simulate structured noise from glandular tissue. The four layers were positioned as a stack and oriented as such that an overlap was created between the masses as well as the microcalcifications. Images were acquired by digital mammography (GE Healthcare Essential) and digital breast tomosynthesis (DBT) on a prototype system (GE Healthcare).

Attenuation experiments on the two rubber compounds showed linear attenuation coefficients of 1,23 cm-1 (base) and 1,27 cm-1 (masses) respectively. These are somewhat higher than tabulated values for ICRU-44 breast tissue (1,11 cm-1), but still in adequate agreement. Standard digital mammography of the phantom proved difficult to read. Of all 24 masses, only the five largest could be differentiated. The clusters of microcalcifications were detectable but heavily masked by overlying tissue. A DBT study with various acquisition setups managed to detect virtually all simulated masses, whether overlapping or not. The microcalcifications were clearly visualised and suffered less from overlying tissue. The relatively sharp edges of both the phantom and the pieces of udder caused artefacts on the reconstructed DBT images. A conic-shaped model should probably overcome this problem. Also our model contained some small air cavities, which caused artefacts as well. Both flaws can be easily remediated in a future model. A 6-month radiographic follow up of the phantom showed no signs of degradation or dehydration.

The constructed phantom seems a valuable tool for DBT quality assessment and acquisition optimization.

The study can be used for acquisition optimization in DBT.

Raeymaeckers, S, Buls, N, Souchay, H, De Mey, J, Breucq, C, Image Quality Phantom for Digital Breast Tomosynthesis.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8015784.html