Background
Recently, the ordered subset expectation maximization (OSEM) algorithm was developed to reconstruct the PET and SPECT images. This iterative algorithm allowed incorporation of realistic modeling of the data acquisition process and statistical noise, which refined estimates of the activity distribution. Since the statistical image analyses such as the Statistical Parametric (SPM) or three dimensional stereotactic surface display (3D-SSP) methods rely on the accurate measurement of activity distribution in the brain, the results may be affected by the use of the OSEM. In this study, we show that the image reconstruction algorithm employed affects the result of the statistical image analysis in mild Alzheimer's disease.
Materials and Methods
Sixteen patients with probable Alzheimer's disease (8 men and 8 women; age range, 56–78 y; mean age, 70.6 years) and 6 normal controls (1 men and 5 women; age range, 48–69; mean age, 62.8 years) were studied. Their diagnosis was based on the criteria of the National Institute of Neurological and Communicative Disorders and Strokes and the Alzheimer's Diseases and Related Disorders association (NINCDS-ADRDA). All patients underwent neuropsychological testing that revealed quantified, objective evidence of memory impairment with no apparent loss in general cognitive, behavioral, or functional status at the study. The score was greater than 20 on the mini-mental State Examination (MMSE). Cerebral blood flow was measured by means of Tc-99 m HMPAO and a high-performance, 4-head rotating gamma camera equipped with a low-energy, general-purpose, parallel-hole collimator with a spatial resolution of 13.0-mm full width at half-maximum (Gamma View SPECT 2000 H, Hitachi Medical Co). The transaxial images were reconstructed with the Filtered Back Projection algorithm (FBP) and the OSEM (iteration 4; subset 8) in each patient. The SPM99 and 3D-SSP were used for the statistical image analysis. All reconstructed images were smoothed using a Gaussian filter with a full-width half maximum of 12 mm in the SPM analysis. The gray matter threshold was set to be 0.8.
Results
In the SMP analysis, the cluster of significant CBF reduction was found in the lateral temporal cortices and posterior cingulated gyrus (p<0.01) when the FBP was employed. When the OSEM was used, it was found in the posterior cingulated gyrus (p<0.01). In the 3D-SSP analysis, posterior cingulated gyrus hypoperfusion was detected by both reconstruction algorithms.
Discussion
We demonstrated that the result of the SPM analysis is affected by the choice of image reconstruction algorithm. We visually inspected images reconstructed with the OSEM and the FBP. We found that the contrast between the posterior cingulated gyrus and surrounding structures were less in the OSEM than in the FBP. This may be due to the dependence of convergence time on the spatial frequencies in the OSEM procedure.
Conclusion
Although the OSEM algorithm is now the most widely used iterative reconstruction method, it may affect the SPM analysis when mild AD patients are studied.