Effect of scatter underestimation on SUV uptake using the DIXON attenuation map in PET/MRI

Background: PET data correction is an essential step toward accurate tracer uptake quantification. In PET/MRI scanners, the absence or lack of accuracy of bone structures identification in attenuation in the MR attenuation map leads to an underestimation of scatter and overestimation of activity uptake. This effect is expected to be more prominent in anatomical regions like the Pelvis. The aim of this study is to investigate the effect of the absence of bone structures in the DIXON based PET attenuation correction on lesion SUV uptake using synthetic lesions.

Methods: Whole-body FDG patient dataset acquired with the Siemens Biograph PET/MRI scanner. Using the available mMR attenuation correction (MRAC) approaches, the two-point DIXON and ATLAS (including bone). A patient CT attenuation map was registered was to the MRAC using affine transformation. A 20 mm diameter synthetic lesions were inserted across the pelvis region, and PET images were reconstructed with the three attenuation maps. This procedure was repeated in an axially distant slice (figure 1 left-top corner). MRAC to CTAC SUV lesion bias was calculated (Figure 1). During PET image reconstruction, scatter profile were also generated for each MRAC approach and displayed against the CTAC generated scatter profile (Figure 2).

Results: Figure 1 present a global view about MRAC AC to CTAC bias in the whole PET reconstructed image using the ATLAS (top row) and two-point DIXON (middle row) attenuation maps. Figure 1 shows that an image profile passing through MRAC to CTAC SUV bias generated with the DIXON umap depicts the positive bias or SUV overestimation in regions situated between patient femur bones. The pic bias at the edge of the patient pelvis is due to the segmentation and the deletion of the background activity. In Figure 2, scatter profiles generated by the two-point Dixon attenuation map showed a slight underestimation compared to the CTAC generated scatter profile, especially on the femur bones. These differences are translated into a few percent, ~5%, of SUV uptake overestimation. In Figure 3, lesions inserted cross the pelvis showed similar expected behavior, an overestimation of SUV uptake using the two-point DIXON umap. However, results obtained using lesions are more precise and regions specific. The observed behavior for the two-point DIXON attenuation was not observed using the ATLAS attenuation map, including bones.

Conclusions: The use of the two-point DIXON attenuation map for PET/MRI attenuation correction showed and overestimation of the SUV uptake in areas between bone femur due to the absence of bones tissue in the DIXON attenuation map. This behavior was not observed using the ATLAS attenuation map, which includes bone.

• Download figure
• Open in new tab
• Download powerpoint