The effect of inaccurate bone attenuation coefficient and segmentation on reconstructed PET images

Combined PET/MRI scanners will require the use of MR images to attenuation correct the PET acquisition. A recognised issue with this is the lack of bone signal in conventional MR images. One approach is to segment the MRI and use a constant attenuation coefficient in place of the bone. The aim of this study was to investigate the effect of replacing the attenuation coefficients of the bone with various constant values on PET.

Bone segmentation was performed on the computed tomography (CT) components of PET/CT images from nine patients, and the bone replaced with three patient-specific values and three generic values. Attenuation-corrected PET images were reconstructed using the modified CT data and compared with the PET images reconstructed using the original CT data. The resulting effects on regions of interest measurements and in all chest voxels were evaluated.

Replacing the bone with the mean bone value (patient specific) produced the least error with, on an average, a maximum error in the lung of 5%. Of the generic bone replacement values tested, an overestimated bone volume with soft tissue values produced the lowest error with, on an average, up to 36, 20 and 10% error in the bone, soft tissue and lung, respectively.

When the bones are substituted with attenuation coefficients that are higher than the soft-tissue, variations in bone classification can significantly degrade the PET images. Using an overestimated bone volume consisting of soft tissue equivalent attenuation coefficients presents a simple, robust method that is less sensitive to segmentation errors.