MR based attenuation correction and µ map production in PET/MR breast imaging.

Objectives: Attenuation correction (AC) is necessary for data quantification in PET imaging. PET reconstructions without AC appear to show areas of low attenuation (patient surface and lungs) as having increased uptake. A transmission-based AC map (µ-map), obtained from rotating rod sources or through CT, is created and utilized during the PET emission data reconstruction. Traditional methods for µ-map creation are unavailable in PET/MR imaging, therefore, MR imaging sequences (DIXON) are utilized to generate the required µ-map. Voxel intensities in an MR image decrease as their distance from the radio frequency coil increases. This can lead to the creation of incomplete µ-maps. Initial investigation of breast PET/MR imaging with a Siemens Biograph mMR has inspired this discussion.

Methods: Diagnostic MR imaging of the breasts is generally performed with only a dedicated breast coil. A volunteer was initially imaged utilizing a DIXON sequence with only a breast coil. A second scan was then completed utilizing both the breast coil and a surface coil on the subject’s back. To simulate lesions in a patient, conical 1.5 mL vials containing a clinically relevant concentration of F-18 FDG were placed in the axillary and mediastinal regions of the volunteer. Standardized uptake values (SUVs) were obtained from regions of interest drawn around the axillary and mediastinal sources to compare the effects on PET AC images derived from an incomplete µ-map (breast coil only) versus a complete µ-map (breast and surface coils).

Results: The DIXON sequence with the breast coil alone generated an incomplete µ-map, as anticipated. SUVs of the axillary source increased from 2.08 to 10.48 when a complete µ-map was generated with the addition of a surface coil. SUVs of the mediastinal source increased from 1.30 to 3.30.

Conclusions: In PET/MR imaging, care must be taken to ensure that the DIXON sequence has adequate MR signal to generate complete µ-maps. Incomplete µ-maps result in PET reconstructions having uncompensated PET counting losses, inaccurate quantification, and underestimated SUVs.