Template-based attenuation correction of PET in hybrid MR-PET scanners

Abstract

708

Objectives: The present design of hybrid MR-PET does not offer measured attenuation correction (AC) using a transmission scan (PET-based AC = PBA). This problem may be solved by deriving attenuation maps from corresponding anatomic MR images. Alternatively, standardized templates of attenuation maps (template-based AC = TBA) may be applied. This work presents our implementation of the TBA and compares its results with those of PBA.

Methods: We used MR and PET-FDG images of ten normal volunteers (5 male, 5 female). To obtain the attenuation template (AT) the individual PET transmission images were normalized with SPM2 to the transmission-template of the SPM2 package and averaged. To apply the TBA, the MR-T1 template of SPM2 was adapted to the individual MR-T1 image which was registered to the corresponding individual transmission image. The individualized attenuation map was obtained by applying the deformation matrix to the AT and adding the head holder’s attenuation data. The final attenuation map was forward projected to get attenuation factors used for the direct AC of the emission data.

Results: The ten PET images corrected for attenuation with PBA or TBA were compared using regions of interest (ROIs) placed at cortical (frontal, temporal, occipital) and sub-cortical (caudate nucleus (cn), putamen, thalamus) areas. Compared to PBA the ROI values obtained with TBA ranged from -10% to +4%. There were no regional preferences, which is also documented by the large standard deviations: The averaged (n=10) ROI values were between -6.3% ± 2.2% (temporal) and -1.4% ± 3.1% (cn).

Conclusions: TBA allows an acceptable AC for brain PET studies in hybrid MR-PET scanners with lacking transmission facility. Differences from pure male or female AT are under investigation. Routine applications should be independent of SPM.