3D Printed Mammography Image Quality Phantoms with Structured Backgrounds

In this technical feasibility thesis, two concept designs of 3D-printed mammography phantoms for Full Field Digital Mammography (FFDM) and Digital Breast Tomosynthesis (DBT) are presented. Each assembled phantom consists of individual anthropomorphic plates that simulate anatomic noise during imaging. Both phantoms were designed with the CAD software Fusion by Autodesk [1] and Blender 4.2 [2], and realized via Fused Deposition Modelling (FDM). Concept #1, the three component basic model (CBM), consists of three equally thick plates, printed with Poly-Lactic-Acid (PLA) filament. When assembled they resemble a compressed breast during mammography. Each of the plates holds a complicated 3D finely branched tubular inside structure that is supposed to imitate the glandular ducts in the parenchymal tissue. The intermediate space is filled with an adipose-tissue imitating material (coconut oil), that has a linear attenuation for X-rays at the common mammography spectra similar to adipose tissue. The CBM phantom is applicable for FFDM and DBT. Concept #2, Phantom 45, simulating a 45mm thick compressed breast, is based on an 8bit orthographic projection of an artificial duct system. The pixel values of the orthographic projection image were used to elevate individual elements of a 101x101 mesh plane in Blender, resulting in a height profile, that resembles the artificial duct system when imaged in 2D Mammography. Phantom 45 was printed with Polyethylen-terephthalat-glycol (PETG). The intermediate space material is pure paraffin wax, also imitating adipose tissue at the common mammography spectra. Phantom 45 is applicable for FFDM only. Both assembled phantoms and the individual plates have been imaged at 27kVp (CBM), and 28kVp (Phantom45) on a Siemens Healthcare [3] Mammomat Revelation (W/Rh). Further a DBT was conducted on the CBM Phantom at 28kVp. The resulting 2D mammography images and reconstructed DBT slices were visually compared to the CAD design renderings and their respective orthographic projection, with focus on pattern-matching and contrast distinction between the used materials. The inside structures of all individual plates were clearly identifiable, including fine features. The assembled CBM FFDM and DBT images could not be compared visually due to the complex anthropomorphic inside structure, and need to be evaluated further with Imaging Software. Phantom 45 showed some printing-related plate issues with PETG, but was overall also of high quality in terms of pattern matching and contrast distinction. In conclusion both design concepts, and the realization via FDM, represent a combination of existing 3D-printed imaging phantom approaches and new implementation ideas for even more realistic phantoms.