A novel partial volume effect correction method which allows for heterogeneous distribution: The potential advantages in the white matter activity estimation on FDG-PET.

Objectives Evaluating detailed activity distribution in the white matter (WM) on [18F]fluorodeoxyglucose (FDG) PET is challenging because removal of the spill-in from the heterogeneous high-uptake in the grey matter (GM) is difficult for the conventional partial volume effect (PVE) correction methods. The assumption of homogeneous activity in target and background regions in the algorithms of most conventional PVE correction methods causes artifacts when the activity distributes heterogeneously within tissues. Then, we have developed a novel PVE correction method for brain PET, named Voxel-wise Anatomical-region-based Non-Uniform Correction (VANUC), which allows for heterogeneous tissue activity. The aim of this study is to examine the potential advantages of our new method in accurate WM activity estimation despite heterogeneous GM activity located adjacent to the WM region.

Methods The VANUC algorithm requires segmented and coregistered MRI and precise measurements of the point spread function of the PET images. This new algorithm assumes that true voxel values in each tissue component are normally distributed random variables within neighborhood region and determines corrected voxel values using a maximum likelihood estimation. We examined effect of heterogeneous activity in the GM on the total activity and distribution in the WM region using simulation PET data. Mean voxel values in GM, WM and the other region were set to 1, 1/3 and 0, respectively, and random heterogeneity was added in GM region (Gaussian distribution, coefficient of variation (CoV) = 0, 0.5, 1). The results of the VANUC method were compared with modified Müller-Gärtner (mMG) method which is a conventional voxel-wise PVE correction method under the assumption of homogeneous background activity. We also applied this new method to the clinical FDG-PET images and compared with mMG method.

Results Both of VANUC and mMG improved the accuracy of total WM activity estimates on simulated PET images (Table). Although heterogeneous activity in GM region influenced the activity distribution in WM region on the PVE-corrected images, VANUC was less susceptible to GM heterogeneity than mMG (Table). Also in PVE correction of clinical FDG-PET images, VANUC corrected the spill-in from the GM uptake effectively. While mMG method resulted in severe artifacts near the boundary with GM region, VANUC corrected the WM activity distribution more homogeneous than mMG.

Conclusions Based on the heterogeneity assumption closer to the clinical situations, our new method provides more accurate WM activity estimates than a conventional PVE correction method. Our new method is expected to be useful for detecting changes of WM metabolism due to aging, neurodegenerative diseases and the other brain diseases.