Abstract
1697
Objectives: The purpose of this study was to investigate artifacts in a voxel-by-voxel statistical result caused by mismatched spatial resolution. We simulated the artifacts using FDG (fluorodeoxyglucose) PET and MRI data. Methods: The subjects were 38 healthy normal volunteers (mean age 59+/-10 years old) underwent brain FDG PET with an ECAT HR scanner (Siemens/CTI) after 40-60 min after the injection of 370 MBq [F-18] FDG. and 3D volumetric MR imaging using a 1.5 T MR scanner, Visart (Toshiba Medical). The reconstructed PET images were spatially normalized with parameters obtained from coregistrated MR image, and smoothed with Gaussian kernel filter 4, 8, 12mm in FWHM. Two sample t-test was performed to compare the data sets of the different smoothness using SPM2. Original PET images were also anatomically standardized with iNeurostat (MediPhysics, Japan) after they were smoothed with the same smoothing filters as above. 3D-SSP Z score mappings were made to compare smoothed data sets. The same processes and statistical analyses as above were applied to smoothed gray matter images that were segmented from the 3D MR images. Results: SPM analyses showed the high t-value in intermediate areas between gray and white matter or between gray matter and cerebral spinal fluid in the group comparisons. In 3D-SSP analyses, Z values (increase in the less smoothed data set) were seen in areas of the primary moto-sensory area, medial temporal, cerebellar hemispheres, and around cerebral ventricles. Z values (decrease in the less smoothed data set) were observed in areas of the parietotemporal association cortices, frontal association cortices, posterior cingulate, striatum, and thalamus. The similar distribution was found in the comparison of simulated FDG uptake images made from the segmented MRI. Conclusions: Mismatched spatial resolution causes artifacts in voxel-based statistical analyses. The artifacts are generated by smoothing effects to high and low values in neighboring brain structures. The distribution of the artifacts might be like AD change in part.
- Society of Nuclear Medicine, Inc.