Anatomical based reconstruction of dynamic brain PET images using N-[11C]-methylpiperidin-4-YL-proprionate ([11C]-PMP) in elderly normals
Nelissen N., Baete K., Vandenbulcke M., Bormans G., Nuyts J., Kinnard V., De Groot T., Van Laere K., Vandenberghe R., Dupont P.
Background and aims: Recently, we have developed a new PET reconstruction algorithm (AMAP) based on a maximum a-posteriori algorithm and using anatomical a-priori information obtained from MRI (Baete et al. 2004a,b). This algorithm corrects for partial volume effects during the reconstruction. The aim of this study was to study the effect on the modelling parameters when using data reconstructed with this new algorithm. Methods: In 10 healthy subjects (6 male, mean age: 67 yr), data were acquired dynamically for one hour (4 x 15s, 4 x 60s, 2 x 150s, 10 x 300s) on a HR+ after the injection of 300 MBq [11C]-PMP. Arterial blood samples were drawn to measure the plasma input function and to determine the metabolite fraction. A T1-weighted volumetric MRI was acquired and segmented according to an optimized VBM approach using a template of elderly normals. Dynamic PET images were reconstructed using AMAP or 3D-FBP (filtered backprojection). Both reconstruction Methods: included corrections for randoms, scatter and attenuation. A correction for small movements during the PET study was applied and the resulting images were then smoothed with a Gaussian kernel of 6 mm FWHM. An irreversible two-tissue compartment model was applied in each voxel with non-linear fitting of the parameters. A constrained value of K1/k2 was determined for each subject on the basis of a whole brain analysis using an unconstrained model. K1 and k3 parametric images were calculated using the data reconstructed with either one of the two Methods and the parametric maps were compared using SPM2 after warping to the MNI space and smoothing with a Gaussian kernel of 12 mm. Results: A high correlation (see table) between the K1 values obtained with the two reconstruction Methods: was found. The correlation for the k3 values was high in cerebellum, striatum and thalamus but was lower in the neocortex. Furthermore, k3 values calculated with the AMAP data were significantly (pFWEcor < 0.05) higher throughout the whole brain compared to the ones determined from the 3D-FBP data. However, no significant difference was found for the K1-maps. Median values (1/min) for K1 and k3 are given in the table. Conclusion: Kinetic modelling of [11C]-PMP reconstructed using AMAP is feasible and the k3 (but not the K1) values are higher compared to 3D-FBP.