Tissue probability-based AC for neurological PET/MR using SPM8

The quantitative accuracy of brain PET/MR has been reported to be reduced compared
to PET/CT due to difficulties related to bone in MR-based attenuation correction (MRAC),
especially in regions near the skull. To reduce these effects, we designed an offline,
tissue probability-based (TPB-AC) using SPM8, which includes soft tissue, air and
bone. Thus, both the increased attenuation due to skull and the reduced attenuation
due to sinuses are taken into account. The method was developed to use the data collected
with the standard anatomical MRAC consisting of 3D T1 FFE, with isotropic voxel size
of 2 mm. Thus, collection of additional sequences or modifications to the standard
attenuation correction sequence are not required. Data from F18-FDG brain studies
performed both with PET/CT and PET/MR was used in AC validation. Seven patients were
reconstructed with the standard attenuation map, TPB-AC and CTAC in PET/MR. Regional
analysis between attenuation corrected PET images was conducted using CTAC as reference.
Using TPB-AC, the relative difference in seven subjects when compared to CTAC in all
regions was -3 %, -2 %, -5 %, 4 %, -3 %, 1 %, 0 %. In clinical AC the difference was
-9 %, -8 %, -9 %, -3 %, -10 %, -5 %, -6 %. Region-wise, the benefit from TPB-AC was
most pronounced in precentral, superior frontal and middle frontal gyrus where underestimation
from clinical AC of: -11 %, -12 % and -12 % was reduced to: -1%, -2 % and -1 % when
using TPB-AC. Tissue probability-based AC was deemed a promising interim method for
PET/MR AC until standardized, commercialized solutions become widely available. To
test the robustness of the method, patients with dental implants or deformed anatomy
would have to be included in the study group.