Extraction of the cerebellar input curve with factor analysis in C-11 DASB PET brain studies

Objectives Many quantitative methods for analysis of dynamic C-11 DASB PET rely on the cerebellum as a reference tissue input function and assume that it has minimal tracer binding in this structure. Factor analysis (FA) and independent component analysis (ICA) has been used to extract time activity curves in dynamic cardiac and tumor studies in addition to creating parametric factor images. FA was applied to dynamic C-11 DASB brain images to reveal the underlying tissue time activity curves and components of the cerebellum.

Methods C-11 DASB was synthesized using a HPLC loop method where [C-11]methyl iodide was swept into the loop coated with the N-Desmethyl-DASB precursor solution. The mixture was purified by RP HPLC and extracted with C18 sep-pak cartridges. 8 controls and 8 patients were injected with 550 MBq of C-11 DASB at initiation of 90 minute dynamic brain imaging. A volumetric region of interest was placed enclosing the entire cerebellum on the images. The number of underlying components of the cerebellum was assumed to be greater than 1. Preprocessing was performed with principal component analysis (PCA). The algorithm identified the apex of the data structure defined in the PCA space (pixels with minimal mixing). An inverse transformation applied to the apex locations generated the underlying cerebellar time activity components.

Results All controls and patients reveal at least two different tissue components within the cerebellum. All factor images and TACs demonstrate significant uptake (binding) of C-11 DASB in the deeper regions of the cerebellum when compared to the more peripheral (cortical) cerebellar regions.

Conclusions There is less cerebellar binding of DASB ligand in the peripheral (cortical) regions as compared to the deeper regions of the cerebellum. Factor analysis may provide a consistent and reproducible method for extraction of the cerebellar reference tissue for tracer kinetic modeling