Anthropomorphic skull phantom using quantitatively accurate bone mimic material

Dharshan Chandramohan; Peng Cao; Misung Han; Hongyu An; John Sunderland; Paul Kinahan; Richard Laforest; Thomas Hope; Peder Larson

Objectives: Quantitative accuracy of PET in PET/MRI studies is hindered, especially in the head, due to challenges in bone identification in current MR-based attenuation correction methods. Bone offers a unique challenge for MR based attenuation correction as it has low MRI signal due to its rapid relaxation rate and low proton density, though it also has the highest attenuation to PET annihilation photons. Recently UTE/ZTE atlas based or machine learning approaches, have been proposed to remedy this problem. Using various doped plasters, 3D printing, and a casting process, we were able to construct a skull phantom for MR that can be used to test novel attenuation correction methods that include bone and are suitable for PET/MRI. Methods: Plaster is a common building material in the form of a dehydrated powder that when mixed with water sets to form the mineral gypsum (CaSO4 2H2O). The similar mineral composition and density compared with bone make it a good candidate for a bone mimic in imaging phantoms. In order to achieve proper MR and attenuation properties we first performed experiments in small vials to determine the MR T1 and T2[asterisk] relaxivity and attenuation of plaster, both undoped and when doped with copper sulfate, manganese chloride (not shown), a gadolinium chelate, and iodine. We then casted undoped and doped plaster in a skull mold to build an anthropomorphic MR-compatible skull phantom. The positive prototype for creating the mold was generated on a LulzBot TAZ 6 fused deposition modeling 3D-printer using PLA plastic. The geometry was derived from the MGH Angel 001 phantom (https://phantoms.martinos.org/MGH_Angel_001). The negative mold was produced by matrix molding using a quick-curing silicone rubber compound (OOMOO 25, Smooth-On, Inc.). The mold was then filled with undoped plaster and plaster doped with 3% CuSO4 by mass. Quantification of T2[asterisk] was done using a 3D ultrashort TE sequence with 32 echo times ranging from 24 to 5000 μsec. MR scans were acquired at 3T with an acquisition matrix of 112x112x114 and 2mm isotropic voxels. Results: In the experiment using doped plaster in small vials, undoped plaster had an average T2[asterisk] of 1364 microseconds, plaster doped with 2.8% copper sulfate had a T2[asterisk] of 414 microseconds, closer to the T2[asterisk] of bone (~300-500 microseconds). Corresponding average CT attenuation values were 1380 HU and 1182 HU for doped and undoped plaster respectively, which are within the range of typical HU values observed in humans. Using the cast skull models we obtained high quality UTE MR Images. Undoped plaster had an average T2[asterisk] of 1132 microseconds, and 3% copper sulfate doped-plaster had shortened the average T2[asterisk] to 858 microseconds. Conclusions: Plaster doped with copper sulfate is a promising candidate for a bone mimic in PET/MRI phantoms with appropriate MR and attenuation values and potential to be cast into an anthropomorphic phantom shape. Differences in the absolute quantitative values of T2[asterisk] between the cast plaster phantom and the vial experiment may be due to differences in uniformity and composition of the final plaster, also differences in the drying and crystallization process. With some minor refinement of the phantom casting process and adjustment and recalibration of the doping this process can be used to construct a quantitatively accurate plaster phantom to mimic bone in MRI over multiple scanning paradigms, including for evaluation of MR-based attenuation correction methods.

Anthropomorphic skull phantom using quantitatively accurate bone mimic material

View Online

Abstract

Details

Metrics