Direct 4D parametric image estimation in reversible tracer binding imaging

Objectives To investigate image qualities obtained via novel direct 4D parametric image reconstruction algorithms in reversible tracer binding imaging.

Methods Based on a consistent graphical analysis method (Zhou et al., 2009, Neuroimage), two novel 4D reconstruction algorithms were developed to estimate distribution volume images: (i) a fast technique, first creating 'parametric sinograms' from the dynamic datasets followed by a single (e.g. FBP, EM) reconstruction; (ii) an accurate closed-form EM technique, extending the system matrix to include relation between data and parametric image, and using a novel hidden complete-data formulation. The means of parameters estimated from 55 human [11C]raclopride dynamic PET were used for simulation (22 realizations) using a mathematical brain phantom. Images were reconstructed using standard FBP or EM methods followed by modeling, as well as proposed direct methods. Noise vs. bias quantitative measurements were performed in eight regions of the brain (grey, caudate, putamen, thalamus, corpus callosum, white, nucleus accumbens, cerebellum).

Results All EM methods (standard, proposed) outperformed FBP approach, as quantified using noise vs. bias measurements in individual ROIs and overall. Fast technique obtained identical (for FBP) and slightly degraded (for EM) performances, the latter being the case as parametric sinograms are not exactly Poisson-distributed, while it achieved ~1 order of magnitude speed-up in the reconstruction task. Accurate direct 4D EM reconstruction resulted in substantial visual and quantitative accuracy improvements (over 100% noise reduction from 58% to 27%; matched bias).

Conclusions Direct 4D parametric imaging can be used to significantly improve reconstruction speed or accuracy in reversible tracer binding imaging.