Spatial normalization and quantification approaches of PET imaging for neurological disorders

Teng Zhang,Shuang Wu,Xiaohui Zhang,Yiwu Dai,Anxin Wang,Hong Zhang,Mei Tian

DOI: https://doi.org/10.1007/s00259-022-05809-6

2022-01-01

Abstract:Quantification approaches of positron emission tomography (PET) imaging provide user-independent evaluation of pathophysiological processes in living brains, which have been strongly recommended in clinical diagnosis of neurological disorders. Most PET quantification approaches depend on spatial normalization of PET images to brain template; however, the spatial normalization and quantification approaches have not been comprehensively reviewed. In this review, we introduced and compared PET template-based and magnetic resonance imaging (MRI)-aided spatial normalization approaches. Tracer-specific and age-specific PET brain templates were surveyed between 1999 and 2021 for 18 F-FDG, 11 C-PIB, 18 F-Florbetapir, 18 F-THK5317, and etc., as well as adaptive PET template methods. Spatial normalization-based PET quantification approaches were reviewed, including region-of-interest (ROI)-based and voxel-wise quantitative methods. Spatial normalization-based ROI segmentation approaches were introduced, including manual delineation on template, atlas-based segmentation, and multi-atlas approach. Voxel-wise quantification approaches were reviewed, including voxel-wise statistics and principal component analysis. Certain concerns and representative examples of clinical applications were provided for both ROI-based and voxel-wise quantification approaches. At last, a recipe for PET spatial normalization and quantification approaches was concluded to improve diagnosis accuracy of neurological disorders in clinical practice.