Comparison of Count Normalization Methods for Statistical Parametric Mapping Analysis Using a Digital Brain Phantom Obtained from Fluorodeoxyglucose-positron Emission Tomography

Alternative normalization methods were proposed to solve the biased information of SPM in the study of neurodegenerative disease. The objective of this study was to determine the most suitable count normalization method for SPM analysis of a neurodegenerative disease based on the results of different count normalization methods applied on a prepared digital phantom similar to one obtained using fluorodeoxyglucose-positron emission tomography (FDG-PET) data of a brain with a known neurodegenerative condition. Digital brain phantoms, mimicking mild and intermediate neurodegenerative disease conditions, were prepared from the FDG-PET data of 11 healthy subjects. SPM analysis was performed on these simulations using different count normalization methods. In the slight-decrease phantom simulation, the Yakushev method correctly visualized wider areas of slightly decreased metabolism with the smallest artifacts of increased metabolism. Other count normalization methods were unable to identify this slightly decreases and produced more artifacts. The intermediate-decreased areas were well visualized by all methods. The areas surrounding the grey matter with the slight decreases were not visualized with the GM and VOI count normalization methods but with the Andersson. The Yakushev method well visualized these areas. Artifacts were present in all methods. When the number of reference area extraction was increased, the Andersson method better-captured the areas with decreased metabolism and reduced the artifacts of increased metabolism. In the Yakushev method, increasing the threshold for the reference area extraction reduced such artifacts. The Yakushev method is the most suitable count normalization method for the SPM analysis of neurodegenerative disease.